WEBVTT Kind: captions Language: en 00:00:00.000 --> 00:00:01.429 align:start position:0% It's<00:00:00.240> no<00:00:00.400> secret<00:00:00.640> that<00:00:00.800> Cursor<00:00:01.120> has<00:00:01.280> been 00:00:01.429 --> 00:00:01.439 align:start position:0% It's no secret that Cursor has been 00:00:01.439 --> 00:00:03.030 align:start position:0% It's no secret that Cursor has been trying<00:00:01.600> to<00:00:01.680> cook<00:00:01.920> their<00:00:02.159> own<00:00:02.320> model<00:00:02.560> for<00:00:02.720> a<00:00:02.879> bit 00:00:03.030 --> 00:00:03.040 align:start position:0% trying to cook their own model for a bit 00:00:03.040 --> 00:00:04.630 align:start position:0% trying to cook their own model for a bit now,<00:00:03.280> and<00:00:03.439> it<00:00:03.600> seems<00:00:03.760> like<00:00:03.919> they<00:00:04.080> succeeded. 00:00:04.630 --> 00:00:04.640 align:start position:0% now, and it seems like they succeeded. 00:00:04.640 --> 00:00:06.789 align:start position:0% now, and it seems like they succeeded. Composer<00:00:05.120> 2<00:00:05.359> is<00:00:05.600> incredible.<00:00:06.080> It<00:00:06.319> beats<00:00:06.560> out 00:00:06.789 --> 00:00:06.799 align:start position:0% Composer 2 is incredible. It beats out 00:00:06.799 --> 00:00:08.310 align:start position:0% Composer 2 is incredible. It beats out Opus<00:00:07.200> in<00:00:07.359> a<00:00:07.520> lot<00:00:07.600> of<00:00:07.759> different<00:00:07.919> things. 00:00:08.310 --> 00:00:08.320 align:start position:0% Opus in a lot of different things. 00:00:08.320 --> 00:00:09.750 align:start position:0% Opus in a lot of different things. That's<00:00:08.480> why<00:00:08.639> I'm<00:00:08.800> so<00:00:09.040> excited<00:00:09.200> to<00:00:09.440> show<00:00:09.599> you 00:00:09.750 --> 00:00:09.760 align:start position:0% That's why I'm so excited to show you 00:00:09.760 --> 00:00:12.950 align:start position:0% That's why I'm so excited to show you guys.<00:00:10.880> That's<00:00:11.280> not<00:00:11.519> the<00:00:11.840> new<00:00:12.080> cursor<00:00:12.559> model. 00:00:12.950 --> 00:00:12.960 align:start position:0% guys. That's not the new cursor model. 00:00:12.960 --> 00:00:15.749 align:start position:0% guys. That's not the new cursor model. That's<00:00:13.200> Kimmy<00:00:13.679> K2.5,<00:00:14.639> the<00:00:14.960> openweight<00:00:15.519> model 00:00:15.749 --> 00:00:15.759 align:start position:0% That's Kimmy K2.5, the openweight model 00:00:15.759 --> 00:00:18.230 align:start position:0% That's Kimmy K2.5, the openweight model from<00:00:16.000> Moonshot<00:00:16.560> AI.<00:00:17.119> Why<00:00:17.279> is<00:00:17.440> that<00:00:17.680> here?<00:00:18.080> I 00:00:18.230 --> 00:00:18.240 align:start position:0% from Moonshot AI. Why is that here? I 00:00:18.240 --> 00:00:18.870 align:start position:0% from Moonshot AI. Why is that here? I thought<00:00:18.320> we<00:00:18.480> were<00:00:18.560> supposed<00:00:18.640> to<00:00:18.720> be<00:00:18.800> talking 00:00:18.870 --> 00:00:18.880 align:start position:0% thought we were supposed to be talking 00:00:18.880 --> 00:00:20.470 align:start position:0% thought we were supposed to be talking about<00:00:19.039> Composer<00:00:19.439> 2,<00:00:19.600> the<00:00:19.760> new<00:00:19.840> model<00:00:20.080> cursor 00:00:20.470 --> 00:00:20.480 align:start position:0% about Composer 2, the new model cursor 00:00:20.480 --> 00:00:21.750 align:start position:0% about Composer 2, the new model cursor just<00:00:20.640> shipped<00:00:20.880> that<00:00:21.119> is<00:00:21.279> performing 00:00:21.750 --> 00:00:21.760 align:start position:0% just shipped that is performing 00:00:21.760 --> 00:00:23.990 align:start position:0% just shipped that is performing incredibly<00:00:22.240> well<00:00:22.480> and<00:00:22.720> for<00:00:22.960> insanely<00:00:23.439> cheap. 00:00:23.990 --> 00:00:24.000 align:start position:0% incredibly well and for insanely cheap. 00:00:24.000 --> 00:00:25.590 align:start position:0% incredibly well and for insanely cheap. Well,<00:00:24.560> that's<00:00:24.800> what<00:00:24.880> we<00:00:25.039> were<00:00:25.199> supposed<00:00:25.439> to 00:00:25.590 --> 00:00:25.600 align:start position:0% Well, that's what we were supposed to 00:00:25.600 --> 00:00:27.429 align:start position:0% Well, that's what we were supposed to talk<00:00:25.840> about<00:00:26.160> until<00:00:26.480> this<00:00:26.720> tweet<00:00:26.960> occurred 00:00:27.429 --> 00:00:27.439 align:start position:0% talk about until this tweet occurred 00:00:27.439 --> 00:00:29.109 align:start position:0% talk about until this tweet occurred where<00:00:27.680> somebody<00:00:28.080> happened<00:00:28.400> to<00:00:28.640> notice<00:00:28.880> when 00:00:29.109 --> 00:00:29.119 align:start position:0% where somebody happened to notice when 00:00:29.119 --> 00:00:31.669 align:start position:0% where somebody happened to notice when they<00:00:29.279> were<00:00:29.439> playing<00:00:29.760> with<00:00:30.000> the<00:00:30.320> URLs<00:00:30.960> for 00:00:31.669 --> 00:00:31.679 align:start position:0% they were playing with the URLs for 00:00:31.679 --> 00:00:33.510 align:start position:0% they were playing with the URLs for hitting<00:00:32.000> things<00:00:32.160> through<00:00:32.320> the<00:00:32.480> cursor<00:00:32.880> API 00:00:33.510 --> 00:00:33.520 align:start position:0% hitting things through the cursor API 00:00:33.520 --> 00:00:35.350 align:start position:0% hitting things through the cursor API that<00:00:33.760> the<00:00:33.920> new<00:00:34.079> model<00:00:34.559> had<00:00:34.719> an<00:00:34.960> interesting 00:00:35.350 --> 00:00:35.360 align:start position:0% that the new model had an interesting 00:00:35.360 --> 00:00:37.590 align:start position:0% that the new model had an interesting name<00:00:35.520> in<00:00:35.760> it.<00:00:36.079> Counts<00:00:36.480> any<00:00:36.719> sphere<00:00:37.040> models 00:00:37.590 --> 00:00:37.600 align:start position:0% name in it. Counts any sphere models 00:00:37.600 --> 00:00:44.389 align:start position:0% name in it. Counts any sphere models Kimmy<00:00:38.079> K2.5<00:00:39.440> RL<00:00:40.000> 0317<00:00:41.280> S515<00:00:42.719> fast.<00:00:43.440> Huh.<00:00:44.079> Turns 00:00:44.389 --> 00:00:44.399 align:start position:0% Kimmy K2.5 RL 0317 S515 fast. Huh. Turns 00:00:44.399 --> 00:00:46.869 align:start position:0% Kimmy K2.5 RL 0317 S515 fast. Huh. Turns out<00:00:44.640> Cursor's<00:00:45.120> new<00:00:45.280> model<00:00:45.760> might<00:00:46.079> not<00:00:46.239> be<00:00:46.719> all 00:00:46.869 --> 00:00:46.879 align:start position:0% out Cursor's new model might not be all 00:00:46.879 --> 00:00:48.389 align:start position:0% out Cursor's new model might not be all that<00:00:47.120> new.<00:00:47.520> There's<00:00:47.680> a<00:00:47.760> lot<00:00:47.840> to<00:00:48.000> dig<00:00:48.160> into 00:00:48.389 --> 00:00:48.399 align:start position:0% that new. There's a lot to dig into 00:00:48.399 --> 00:00:50.470 align:start position:0% that new. There's a lot to dig into here.<00:00:48.640> From<00:00:48.879> the<00:00:49.039> actual<00:00:49.440> performance<00:00:50.079> costs 00:00:50.470 --> 00:00:50.480 align:start position:0% here. From the actual performance costs 00:00:50.480 --> 00:00:52.069 align:start position:0% here. From the actual performance costs and<00:00:50.800> experience<00:00:51.200> you<00:00:51.440> can<00:00:51.600> expect<00:00:51.840> from 00:00:52.069 --> 00:00:52.079 align:start position:0% and experience you can expect from 00:00:52.079 --> 00:00:54.389 align:start position:0% and experience you can expect from composer<00:00:52.719> to<00:00:52.960> the<00:00:53.120> weird<00:00:53.600> building<00:00:53.920> on<00:00:54.079> top<00:00:54.239> of 00:00:54.389 --> 00:00:54.399 align:start position:0% composer to the weird building on top of 00:00:54.399 --> 00:00:56.389 align:start position:0% composer to the weird building on top of Kimmy<00:00:54.800> that<00:00:55.039> was<00:00:55.199> not<00:00:55.360> really<00:00:55.680> disclosed<00:00:56.079> and 00:00:56.389 --> 00:00:56.399 align:start position:0% Kimmy that was not really disclosed and 00:00:56.399 --> 00:00:58.069 align:start position:0% Kimmy that was not really disclosed and might<00:00:56.640> be<00:00:56.719> against<00:00:57.039> their<00:00:57.199> license<00:00:57.680> to<00:00:57.920> the 00:00:58.069 --> 00:00:58.079 align:start position:0% might be against their license to the 00:00:58.079 --> 00:01:00.069 align:start position:0% might be against their license to the deleted<00:00:58.480> tweets<00:00:58.879> from<00:00:59.120> members<00:00:59.440> of<00:00:59.520> the<00:00:59.760> Kimmy 00:01:00.069 --> 00:01:00.079 align:start position:0% deleted tweets from members of the Kimmy 00:01:00.079 --> 00:01:02.470 align:start position:0% deleted tweets from members of the Kimmy moonshot<00:01:00.719> team<00:01:01.280> posting<00:01:01.680> about<00:01:02.000> this 00:01:02.470 --> 00:01:02.480 align:start position:0% moonshot team posting about this 00:01:02.480 --> 00:01:05.270 align:start position:0% moonshot team posting about this confused<00:01:03.120> questioning<00:01:03.680> things<00:01:04.320> and<00:01:04.879> a<00:01:05.119> hell 00:01:05.270 --> 00:01:05.280 align:start position:0% confused questioning things and a hell 00:01:05.280 --> 00:01:07.030 align:start position:0% confused questioning things and a hell of<00:01:05.360> a<00:01:05.519> lot<00:01:05.680> more.<00:01:06.320> This<00:01:06.560> one's<00:01:06.720> been<00:01:06.880> a 00:01:07.030 --> 00:01:07.040 align:start position:0% of a lot more. This one's been a 00:01:07.040 --> 00:01:08.390 align:start position:0% of a lot more. This one's been a fascinating<00:01:07.439> deep<00:01:07.680> dive<00:01:07.840> and<00:01:08.000> I've<00:01:08.159> done<00:01:08.240> a 00:01:08.390 --> 00:01:08.400 align:start position:0% fascinating deep dive and I've done a 00:01:08.400 --> 00:01:09.830 align:start position:0% fascinating deep dive and I've done a bunch<00:01:08.560> of<00:01:08.640> research<00:01:08.880> ahead<00:01:09.200> of<00:01:09.360> time<00:01:09.439> to<00:01:09.680> try 00:01:09.830 --> 00:01:09.840 align:start position:0% bunch of research ahead of time to try 00:01:09.840 --> 00:01:11.990 align:start position:0% bunch of research ahead of time to try and<00:01:10.080> cover<00:01:10.240> it<00:01:10.479> to<00:01:10.640> the<00:01:10.799> best<00:01:11.040> of<00:01:11.200> my<00:01:11.360> ability. 00:01:11.990 --> 00:01:12.000 align:start position:0% and cover it to the best of my ability. 00:01:12.000 --> 00:01:13.750 align:start position:0% and cover it to the best of my ability. As<00:01:12.240> you<00:01:12.320> all<00:01:12.479> know,<00:01:12.720> I<00:01:12.960> am<00:01:13.119> an<00:01:13.280> investor<00:01:13.600> in 00:01:13.750 --> 00:01:13.760 align:start position:0% As you all know, I am an investor in 00:01:13.760 --> 00:01:15.429 align:start position:0% As you all know, I am an investor in Cursor<00:01:14.240> but<00:01:14.320> they<00:01:14.479> are<00:01:14.640> not<00:01:14.720> paying<00:01:14.960> me<00:01:15.119> in<00:01:15.280> any 00:01:15.429 --> 00:01:15.439 align:start position:0% Cursor but they are not paying me in any 00:01:15.439 --> 00:01:17.109 align:start position:0% Cursor but they are not paying me in any way,<00:01:15.600> shape<00:01:15.760> or<00:01:15.920> form<00:01:16.159> to<00:01:16.320> cover<00:01:16.479> any<00:01:16.640> of<00:01:16.799> this. 00:01:17.109 --> 00:01:17.119 align:start position:0% way, shape or form to cover any of this. 00:01:17.119 --> 00:01:18.390 align:start position:0% way, shape or form to cover any of this. As<00:01:17.360> you<00:01:17.520> can<00:01:17.680> imagine,<00:01:17.920> they<00:01:18.159> probably<00:01:18.320> don't 00:01:18.390 --> 00:01:18.400 align:start position:0% As you can imagine, they probably don't 00:01:18.400 --> 00:01:20.070 align:start position:0% As you can imagine, they probably don't want<00:01:18.560> me<00:01:18.720> covering<00:01:18.880> this<00:01:19.119> drama<00:01:19.439> too<00:01:19.600> much. 00:01:20.070 --> 00:01:20.080 align:start position:0% want me covering this drama too much. 00:01:20.080 --> 00:01:21.910 align:start position:0% want me covering this drama too much. Never<00:01:20.400> been<00:01:20.560> the<00:01:20.720> biggest<00:01:20.960> fan<00:01:21.119> of<00:01:21.280> drama,<00:01:21.680> but 00:01:21.910 --> 00:01:21.920 align:start position:0% Never been the biggest fan of drama, but 00:01:21.920 --> 00:01:23.270 align:start position:0% Never been the biggest fan of drama, but this<00:01:22.080> one's<00:01:22.400> important<00:01:22.640> enough<00:01:22.880> and<00:01:23.119> has 00:01:23.270 --> 00:01:23.280 align:start position:0% this one's important enough and has 00:01:23.280 --> 00:01:24.710 align:start position:0% this one's important enough and has enough<00:01:23.520> good<00:01:23.680> learnings<00:01:24.080> that<00:01:24.240> I<00:01:24.400> think<00:01:24.479> it's 00:01:24.710 --> 00:01:24.720 align:start position:0% enough good learnings that I think it's 00:01:24.720 --> 00:01:26.870 align:start position:0% enough good learnings that I think it's worth<00:01:25.040> covering<00:01:25.360> and<00:01:25.600> discussing<00:01:26.000> in<00:01:26.320> detail. 00:01:26.870 --> 00:01:26.880 align:start position:0% worth covering and discussing in detail. 00:01:26.880 --> 00:01:28.390 align:start position:0% worth covering and discussing in detail. There<00:01:27.119> is<00:01:27.280> something<00:01:27.360> I<00:01:27.600> am<00:01:27.680> a<00:01:27.840> huge<00:01:28.000> fan<00:01:28.159> of 00:01:28.390 --> 00:01:28.400 align:start position:0% There is something I am a huge fan of 00:01:28.400 --> 00:01:30.070 align:start position:0% There is something I am a huge fan of though,<00:01:28.799> today's<00:01:29.119> sponsor.<00:01:29.520> Nowadays,<00:01:29.920> it 00:01:30.070 --> 00:01:30.080 align:start position:0% though, today's sponsor. Nowadays, it 00:01:30.080 --> 00:01:31.030 align:start position:0% though, today's sponsor. Nowadays, it feels<00:01:30.320> like<00:01:30.400> there's<00:01:30.640> nothing<00:01:30.799> that 00:01:31.030 --> 00:01:31.040 align:start position:0% feels like there's nothing that 00:01:31.040 --> 00:01:33.190 align:start position:0% feels like there's nothing that Anthropic<00:01:31.600> and<00:01:31.759> OpenAI<00:01:32.400> can<00:01:32.479> agree<00:01:32.720> on.<00:01:33.040> What 00:01:33.190 --> 00:01:33.200 align:start position:0% Anthropic and OpenAI can agree on. What 00:01:33.200 --> 00:01:34.710 align:start position:0% Anthropic and OpenAI can agree on. What if<00:01:33.360> I<00:01:33.439> told<00:01:33.600> you<00:01:33.759> there's<00:01:34.079> one<00:01:34.320> important 00:01:34.710 --> 00:01:34.720 align:start position:0% if I told you there's one important 00:01:34.720 --> 00:01:36.550 align:start position:0% if I told you there's one important thing<00:01:34.880> that<00:01:35.119> they<00:01:35.280> do<00:01:35.439> agree<00:01:35.759> on,<00:01:36.159> as<00:01:36.320> well<00:01:36.400> as 00:01:36.550 --> 00:01:36.560 align:start position:0% thing that they do agree on, as well as 00:01:36.560 --> 00:01:39.190 align:start position:0% thing that they do agree on, as well as T3<00:01:37.040> Chat,<00:01:37.360> Verscell,<00:01:37.920> Perplexity,<00:01:38.560> and<00:01:38.799> many 00:01:39.190 --> 00:01:39.200 align:start position:0% T3 Chat, Verscell, Perplexity, and many 00:01:39.200 --> 00:01:41.429 align:start position:0% T3 Chat, Verscell, Perplexity, and many other<00:01:39.439> businesses.<00:01:40.320> In<00:01:40.560> fact,<00:01:40.799> it's<00:01:40.960> a<00:01:41.200> thing 00:01:41.429 --> 00:01:41.439 align:start position:0% other businesses. In fact, it's a thing 00:01:41.439 --> 00:01:42.950 align:start position:0% other businesses. In fact, it's a thing all<00:01:41.680> of<00:01:41.840> these<00:01:42.159> companies<00:01:42.479> agree<00:01:42.720> on, 00:01:42.950 --> 00:01:42.960 align:start position:0% all of these companies agree on, 00:01:42.960 --> 00:01:44.870 align:start position:0% all of these companies agree on, including<00:01:43.280> Carta,<00:01:43.680> Replet,<00:01:44.159> AMP,<00:01:44.560> Bolt, 00:01:44.870 --> 00:01:44.880 align:start position:0% including Carta, Replet, AMP, Bolt, 00:01:44.880 --> 00:01:47.830 align:start position:0% including Carta, Replet, AMP, Bolt, Vanta,<00:01:45.360> and<00:01:45.840> so<00:01:46.240> many<00:01:46.560> other<00:01:46.880> big<00:01:47.119> businesses. 00:01:47.830 --> 00:01:47.840 align:start position:0% Vanta, and so many other big businesses. 00:01:47.840 --> 00:01:48.870 align:start position:0% Vanta, and so many other big businesses. You<00:01:48.000> probably<00:01:48.159> know<00:01:48.320> what<00:01:48.479> it<00:01:48.560> is<00:01:48.640> if<00:01:48.799> you've 00:01:48.870 --> 00:01:48.880 align:start position:0% You probably know what it is if you've 00:01:48.880 --> 00:01:50.550 align:start position:0% You probably know what it is if you've been<00:01:49.040> watching<00:01:49.280> my<00:01:49.439> videos<00:01:49.759> recently.<00:01:50.320> It's 00:01:50.550 --> 00:01:50.560 align:start position:0% been watching my videos recently. It's 00:01:50.560 --> 00:01:52.310 align:start position:0% been watching my videos recently. It's work<00:01:50.880> OS.<00:01:51.280> The<00:01:51.520> best<00:01:51.680> solution<00:01:52.079> for 00:01:52.310 --> 00:01:52.320 align:start position:0% work OS. The best solution for 00:01:52.320 --> 00:01:54.149 align:start position:0% work OS. The best solution for enterprise<00:01:52.880> ready<00:01:53.119> O.<00:01:53.520> If<00:01:53.680> you<00:01:53.759> want<00:01:53.840> to<00:01:54.000> make 00:01:54.149 --> 00:01:54.159 align:start position:0% enterprise ready O. If you want to make 00:01:54.159 --> 00:01:55.990 align:start position:0% enterprise ready O. If you want to make your<00:01:54.479> project<00:01:54.880> ready<00:01:55.119> to<00:01:55.280> go<00:01:55.439> for<00:01:55.680> big 00:01:55.990 --> 00:01:56.000 align:start position:0% your project ready to go for big 00:01:56.000 --> 00:01:57.749 align:start position:0% your project ready to go for big companies<00:01:56.320> to<00:01:56.560> adopt,<00:01:56.960> there<00:01:57.200> is<00:01:57.360> no<00:01:57.520> better 00:01:57.749 --> 00:01:57.759 align:start position:0% companies to adopt, there is no better 00:01:57.759 --> 00:01:59.670 align:start position:0% companies to adopt, there is no better place<00:01:58.000> to<00:01:58.159> start<00:01:58.399> than<00:01:58.640> work<00:01:58.880> OS.<00:01:59.280> There's<00:01:59.520> two 00:01:59.670 --> 00:01:59.680 align:start position:0% place to start than work OS. There's two 00:01:59.680 --> 00:02:01.109 align:start position:0% place to start than work OS. There's two types<00:01:59.840> of<00:02:00.000> devs<00:02:00.320> in<00:02:00.399> the<00:02:00.479> world.<00:02:00.799> There's<00:02:00.960> the 00:02:01.109 --> 00:02:01.119 align:start position:0% types of devs in the world. There's the 00:02:01.119 --> 00:02:03.429 align:start position:0% types of devs in the world. There's the type<00:02:01.280> that<00:02:01.520> hear<00:02:01.680> about<00:02:01.920> Octa,<00:02:02.399> SAML,<00:02:02.960> Duo, 00:02:03.429 --> 00:02:03.439 align:start position:0% type that hear about Octa, SAML, Duo, 00:02:03.439 --> 00:02:06.069 align:start position:0% type that hear about Octa, SAML, Duo, ADP,<00:02:04.159> ADFS,<00:02:04.799> and<00:02:04.960> more<00:02:05.439> and<00:02:05.680> don't<00:02:05.920> really 00:02:06.069 --> 00:02:06.079 align:start position:0% ADP, ADFS, and more and don't really 00:02:06.079 --> 00:02:07.670 align:start position:0% ADP, ADFS, and more and don't really know<00:02:06.240> what<00:02:06.399> they<00:02:06.640> are.<00:02:06.960> And<00:02:07.119> then<00:02:07.280> there's<00:02:07.520> the 00:02:07.670 --> 00:02:07.680 align:start position:0% know what they are. And then there's the 00:02:07.680 --> 00:02:10.070 align:start position:0% know what they are. And then there's the type<00:02:07.840> of<00:02:08.000> dev<00:02:08.319> that<00:02:08.640> feels<00:02:09.039> their<00:02:09.280> heart<00:02:09.679> sink 00:02:10.070 --> 00:02:10.080 align:start position:0% type of dev that feels their heart sink 00:02:10.080 --> 00:02:11.910 align:start position:0% type of dev that feels their heart sink as<00:02:10.239> soon<00:02:10.399> as<00:02:10.640> any<00:02:10.959> of<00:02:11.120> these<00:02:11.440> terms<00:02:11.760> are 00:02:11.910 --> 00:02:11.920 align:start position:0% as soon as any of these terms are 00:02:11.920 --> 00:02:13.589 align:start position:0% as soon as any of these terms are mentioned<00:02:12.400> because<00:02:12.640> setting<00:02:12.959> all<00:02:13.120> of<00:02:13.200> this<00:02:13.440> up 00:02:13.589 --> 00:02:13.599 align:start position:0% mentioned because setting all of this up 00:02:13.599 --> 00:02:15.510 align:start position:0% mentioned because setting all of this up is<00:02:13.920> the<00:02:14.160> worst.<00:02:14.720> And<00:02:14.959> every<00:02:15.200> different 00:02:15.510 --> 00:02:15.520 align:start position:0% is the worst. And every different 00:02:15.520 --> 00:02:16.869 align:start position:0% is the worst. And every different company<00:02:15.920> has<00:02:16.080> their<00:02:16.319> own<00:02:16.480> weird 00:02:16.869 --> 00:02:16.879 align:start position:0% company has their own weird 00:02:16.879 --> 00:02:18.710 align:start position:0% company has their own weird implementation<00:02:17.440> of<00:02:17.760> one<00:02:18.000> of<00:02:18.080> these<00:02:18.400> many 00:02:18.710 --> 00:02:18.720 align:start position:0% implementation of one of these many 00:02:18.720 --> 00:02:20.390 align:start position:0% implementation of one of these many different<00:02:19.040> standards.<00:02:19.680> And<00:02:19.840> if<00:02:20.000> you<00:02:20.080> want<00:02:20.239> the 00:02:20.390 --> 00:02:20.400 align:start position:0% different standards. And if you want the 00:02:20.400 --> 00:02:21.830 align:start position:0% different standards. And if you want the employees<00:02:20.800> of<00:02:20.959> that<00:02:21.120> company<00:02:21.360> to<00:02:21.520> be<00:02:21.680> able<00:02:21.760> to 00:02:21.830 --> 00:02:21.840 align:start position:0% employees of that company to be able to 00:02:21.840 --> 00:02:23.510 align:start position:0% employees of that company to be able to sign<00:02:22.080> into<00:02:22.319> your<00:02:22.640> application,<00:02:23.280> you're 00:02:23.510 --> 00:02:23.520 align:start position:0% sign into your application, you're 00:02:23.520 --> 00:02:25.190 align:start position:0% sign into your application, you're either<00:02:23.760> going<00:02:23.840> to<00:02:23.920> write<00:02:24.160> a<00:02:24.400> lot<00:02:24.480> of<00:02:24.640> miserable 00:02:25.190 --> 00:02:25.200 align:start position:0% either going to write a lot of miserable 00:02:25.200 --> 00:02:27.670 align:start position:0% either going to write a lot of miserable glue<00:02:25.520> code<00:02:25.680> and<00:02:25.840> be<00:02:26.000> the<00:02:26.160> IT<00:02:26.560> team's<00:02:26.959> nightmare 00:02:27.670 --> 00:02:27.680 align:start position:0% glue code and be the IT team's nightmare 00:02:27.680 --> 00:02:29.910 align:start position:0% glue code and be the IT team's nightmare or<00:02:28.000> you're<00:02:28.239> using<00:02:28.480> work<00:02:28.800> OS<00:02:29.200> and<00:02:29.360> you<00:02:29.520> can<00:02:29.680> just 00:02:29.910 --> 00:02:29.920 align:start position:0% or you're using work OS and you can just 00:02:29.920 --> 00:02:31.750 align:start position:0% or you're using work OS and you can just send<00:02:30.160> a<00:02:30.319> link<00:02:30.480> to<00:02:30.640> the<00:02:30.800> IT<00:02:31.120> team<00:02:31.280> at<00:02:31.520> that 00:02:31.750 --> 00:02:31.760 align:start position:0% send a link to the IT team at that 00:02:31.760 --> 00:02:33.430 align:start position:0% send a link to the IT team at that company<00:02:32.000> and<00:02:32.239> they're<00:02:32.480> ready<00:02:32.640> to<00:02:32.879> go.<00:02:33.120> And<00:02:33.280> I 00:02:33.430 --> 00:02:33.440 align:start position:0% company and they're ready to go. And I 00:02:33.440 --> 00:02:34.229 align:start position:0% company and they're ready to go. And I probably<00:02:33.599> should<00:02:33.760> have<00:02:33.920> mentioned<00:02:34.080> this 00:02:34.229 --> 00:02:34.239 align:start position:0% probably should have mentioned this 00:02:34.239 --> 00:02:35.589 align:start position:0% probably should have mentioned this before.<00:02:34.720> It's<00:02:34.879> not<00:02:35.040> hidden<00:02:35.280> in<00:02:35.440> some 00:02:35.589 --> 00:02:35.599 align:start position:0% before. It's not hidden in some 00:02:35.599 --> 00:02:37.990 align:start position:0% before. It's not hidden in some dashboard.<00:02:36.319> You<00:02:36.480> can<00:02:36.640> write<00:02:36.959> code<00:02:37.360> to<00:02:37.599> trigger 00:02:37.990 --> 00:02:38.000 align:start position:0% dashboard. You can write code to trigger 00:02:38.000 --> 00:02:39.350 align:start position:0% dashboard. You can write code to trigger the<00:02:38.160> onboarding.<00:02:38.800> If<00:02:38.959> you're<00:02:39.040> trying<00:02:39.120> to<00:02:39.280> make 00:02:39.350 --> 00:02:39.360 align:start position:0% the onboarding. If you're trying to make 00:02:39.360 --> 00:02:41.110 align:start position:0% the onboarding. If you're trying to make it<00:02:39.599> easier<00:02:39.760> for<00:02:40.000> real<00:02:40.239> businesses<00:02:40.560> to<00:02:40.720> onboard 00:02:41.110 --> 00:02:41.120 align:start position:0% it easier for real businesses to onboard 00:02:41.120 --> 00:02:43.910 align:start position:0% it easier for real businesses to onboard into<00:02:41.360> your<00:02:41.519> app,<00:02:41.840> look<00:02:42.080> no<00:02:42.239> further<00:02:42.400> than<00:02:42.640> soy. 00:02:43.910 --> 00:02:43.920 align:start position:0% into your app, look no further than soy. 00:02:43.920 --> 00:02:46.309 align:start position:0% into your app, look no further than soy. OoS.<00:02:44.560> So,<00:02:44.720> let's<00:02:44.959> start<00:02:45.200> from<00:02:45.519> the<00:02:45.760> top. 00:02:46.309 --> 00:02:46.319 align:start position:0% OoS. So, let's start from the top. 00:02:46.319 --> 00:02:48.869 align:start position:0% OoS. So, let's start from the top. Composer<00:02:47.120> 2.<00:02:47.519> This<00:02:47.680> is<00:02:47.760> a<00:02:48.080> new<00:02:48.319> model<00:02:48.640> that 00:02:48.869 --> 00:02:48.879 align:start position:0% Composer 2. This is a new model that 00:02:48.879 --> 00:02:51.670 align:start position:0% Composer 2. This is a new model that exists<00:02:49.280> incursor<00:02:50.000> and<00:02:50.319> only<00:02:50.720> incursor,<00:02:51.440> which 00:02:51.670 --> 00:02:51.680 align:start position:0% exists incursor and only incursor, which 00:02:51.680 --> 00:02:53.830 align:start position:0% exists incursor and only incursor, which is<00:02:51.760> a<00:02:52.080> a<00:02:52.400> thing<00:02:52.560> I'm<00:02:52.720> not<00:02:52.879> the<00:02:53.040> biggest<00:02:53.360> fan<00:02:53.519> of 00:02:53.830 --> 00:02:53.840 align:start position:0% is a a thing I'm not the biggest fan of 00:02:53.840 --> 00:02:55.910 align:start position:0% is a a thing I'm not the biggest fan of for<00:02:54.080> a<00:02:54.239> bunch<00:02:54.400> of<00:02:54.560> different<00:02:54.879> reasons.<00:02:55.519> I<00:02:55.680> do 00:02:55.910 --> 00:02:55.920 align:start position:0% for a bunch of different reasons. I do 00:02:55.920 --> 00:02:57.110 align:start position:0% for a bunch of different reasons. I do kind<00:02:56.080> of<00:02:56.160> understand<00:02:56.480> why<00:02:56.720> they<00:02:56.879> don't<00:02:56.959> want 00:02:57.110 --> 00:02:57.120 align:start position:0% kind of understand why they don't want 00:02:57.120 --> 00:02:59.509 align:start position:0% kind of understand why they don't want to<00:02:57.280> expose<00:02:57.599> this<00:02:57.840> over<00:02:58.160> API.<00:02:58.879> It's<00:02:59.120> a<00:02:59.280> model 00:02:59.509 --> 00:02:59.519 align:start position:0% to expose this over API. It's a model 00:02:59.519 --> 00:03:01.670 align:start position:0% to expose this over API. It's a model that's<00:02:59.840> only<00:03:00.160> good<00:03:00.400> for<00:03:00.720> code.<00:03:01.120> So<00:03:01.360> even<00:03:01.519> the 00:03:01.670 --> 00:03:01.680 align:start position:0% that's only good for code. So even the 00:03:01.680 --> 00:03:03.430 align:start position:0% that's only good for code. So even the benchmarks<00:03:02.239> are<00:03:02.400> going<00:03:02.480> to<00:03:02.560> be<00:03:02.720> weird<00:03:03.200> because 00:03:03.430 --> 00:03:03.440 align:start position:0% benchmarks are going to be weird because 00:03:03.440 --> 00:03:04.790 align:start position:0% benchmarks are going to be weird because a<00:03:03.599> lot<00:03:03.680> of<00:03:03.760> the<00:03:03.920> benchmarks<00:03:04.319> that<00:03:04.560> exist, 00:03:04.790 --> 00:03:04.800 align:start position:0% a lot of the benchmarks that exist, 00:03:04.800 --> 00:03:06.390 align:start position:0% a lot of the benchmarks that exist, things<00:03:04.959> like<00:03:05.120> artificial<00:03:05.599> analysis,<00:03:06.000> which<00:03:06.239> I 00:03:06.390 --> 00:03:06.400 align:start position:0% things like artificial analysis, which I 00:03:06.400 --> 00:03:08.390 align:start position:0% things like artificial analysis, which I love,<00:03:06.800> have<00:03:06.959> a<00:03:07.200> lot<00:03:07.280> of<00:03:07.440> things<00:03:07.680> in<00:03:07.840> the<00:03:08.000> bench 00:03:08.390 --> 00:03:08.400 align:start position:0% love, have a lot of things in the bench 00:03:08.400 --> 00:03:10.070 align:start position:0% love, have a lot of things in the bench that<00:03:08.640> aren't<00:03:09.040> just<00:03:09.280> testing<00:03:09.680> coding 00:03:10.070 --> 00:03:10.080 align:start position:0% that aren't just testing coding 00:03:10.080 --> 00:03:11.589 align:start position:0% that aren't just testing coding capabilities.<00:03:10.720> And<00:03:10.959> this<00:03:11.120> version<00:03:11.360> of<00:03:11.440> the 00:03:11.589 --> 00:03:11.599 align:start position:0% capabilities. And this version of the 00:03:11.599 --> 00:03:13.910 align:start position:0% capabilities. And this version of the model<00:03:11.920> will<00:03:12.159> probably<00:03:12.560> be<00:03:12.800> very<00:03:13.200> bad<00:03:13.360> at<00:03:13.519> that 00:03:13.910 --> 00:03:13.920 align:start position:0% model will probably be very bad at that 00:03:13.920 --> 00:03:16.149 align:start position:0% model will probably be very bad at that because<00:03:14.239> Composer<00:03:14.959> is<00:03:15.200> meant<00:03:15.440> to<00:03:15.599> be<00:03:15.840> a 00:03:16.149 --> 00:03:16.159 align:start position:0% because Composer is meant to be a 00:03:16.159 --> 00:03:18.309 align:start position:0% because Composer is meant to be a fine-tuned<00:03:16.800> model<00:03:17.040> that<00:03:17.360> does<00:03:17.599> nothing<00:03:18.000> but 00:03:18.309 --> 00:03:18.319 align:start position:0% fine-tuned model that does nothing but 00:03:18.319 --> 00:03:19.750 align:start position:0% fine-tuned model that does nothing but code,<00:03:18.640> which<00:03:18.879> I<00:03:19.040> actually<00:03:19.280> think<00:03:19.360> is<00:03:19.519> cool. 00:03:19.750 --> 00:03:19.760 align:start position:0% code, which I actually think is cool. 00:03:19.760 --> 00:03:21.350 align:start position:0% code, which I actually think is cool. It's<00:03:20.000> nice<00:03:20.080> to<00:03:20.239> see<00:03:20.400> this<00:03:20.640> type<00:03:20.800> of<00:03:21.040> research 00:03:21.350 --> 00:03:21.360 align:start position:0% It's nice to see this type of research 00:03:21.360 --> 00:03:23.270 align:start position:0% It's nice to see this type of research going<00:03:21.599> into<00:03:21.840> a<00:03:22.080> bespoke<00:03:22.560> model<00:03:22.800> that<00:03:23.040> does 00:03:23.270 --> 00:03:23.280 align:start position:0% going into a bespoke model that does 00:03:23.280 --> 00:03:25.190 align:start position:0% going into a bespoke model that does just<00:03:23.599> one<00:03:23.840> thing<00:03:24.000> that<00:03:24.319> matters.<00:03:24.959> But<00:03:25.040> the 00:03:25.190 --> 00:03:25.200 align:start position:0% just one thing that matters. But the 00:03:25.200 --> 00:03:26.790 align:start position:0% just one thing that matters. But the path<00:03:25.360> that<00:03:25.519> they<00:03:25.680> use<00:03:25.840> to<00:03:26.080> get<00:03:26.159> there<00:03:26.400> is<00:03:26.640> where 00:03:26.790 --> 00:03:26.800 align:start position:0% path that they use to get there is where 00:03:26.800 --> 00:03:28.309 align:start position:0% path that they use to get there is where things<00:03:27.040> get<00:03:27.280> really<00:03:27.599> interesting<00:03:28.000> in<00:03:28.159> my 00:03:28.309 --> 00:03:28.319 align:start position:0% things get really interesting in my 00:03:28.319 --> 00:03:29.830 align:start position:0% things get really interesting in my opinion.<00:03:28.800> I've<00:03:29.040> been<00:03:29.200> keeping<00:03:29.360> an<00:03:29.519> eye<00:03:29.599> on<00:03:29.680> the 00:03:29.830 --> 00:03:29.840 align:start position:0% opinion. I've been keeping an eye on the 00:03:29.840 --> 00:03:31.670 align:start position:0% opinion. I've been keeping an eye on the composer<00:03:30.319> models<00:03:30.640> for<00:03:30.799> a<00:03:30.959> while<00:03:31.120> and<00:03:31.360> for<00:03:31.519> a 00:03:31.670 --> 00:03:31.680 align:start position:0% composer models for a while and for a 00:03:31.680 --> 00:03:33.509 align:start position:0% composer models for a while and for a reason<00:03:31.920> that's<00:03:32.319> probably<00:03:32.640> a<00:03:32.959> surprise<00:03:33.200> to<00:03:33.360> you 00:03:33.509 --> 00:03:33.519 align:start position:0% reason that's probably a surprise to you 00:03:33.519 --> 00:03:35.589 align:start position:0% reason that's probably a surprise to you guys.<00:03:33.920> Y'all<00:03:34.159> remember<00:03:34.400> Super<00:03:34.640> Maven?<00:03:35.280> Super 00:03:35.589 --> 00:03:35.599 align:start position:0% guys. Y'all remember Super Maven? Super 00:03:35.599 --> 00:03:38.070 align:start position:0% guys. Y'all remember Super Maven? Super Maven<00:03:36.080> was<00:03:36.319> the<00:03:36.560> first<00:03:36.959> autocomplete<00:03:37.680> AI 00:03:38.070 --> 00:03:38.080 align:start position:0% Maven was the first autocomplete AI 00:03:38.080 --> 00:03:39.750 align:start position:0% Maven was the first autocomplete AI thing<00:03:38.239> that<00:03:38.400> I<00:03:38.560> used<00:03:38.640> in<00:03:38.879> my<00:03:38.959> editor<00:03:39.440> that 00:03:39.750 --> 00:03:39.760 align:start position:0% thing that I used in my editor that 00:03:39.760 --> 00:03:42.070 align:start position:0% thing that I used in my editor that truly<00:03:40.159> blew<00:03:40.400> me<00:03:40.560> away.<00:03:41.040> Copilot<00:03:41.599> had<00:03:41.760> moments 00:03:42.070 --> 00:03:42.080 align:start position:0% truly blew me away. Copilot had moments 00:03:42.080 --> 00:03:43.670 align:start position:0% truly blew me away. Copilot had moments that<00:03:42.239> were<00:03:42.400> really<00:03:42.640> cool,<00:03:42.879> but<00:03:43.120> Super<00:03:43.360> Maven 00:03:43.670 --> 00:03:43.680 align:start position:0% that were really cool, but Super Maven 00:03:43.680 --> 00:03:45.589 align:start position:0% that were really cool, but Super Maven had<00:03:43.840> this<00:03:44.000> weird<00:03:44.239> way<00:03:44.319> of<00:03:44.480> like<00:03:44.799> learning<00:03:45.360> as 00:03:45.589 --> 00:03:45.599 align:start position:0% had this weird way of like learning as 00:03:45.599 --> 00:03:47.830 align:start position:0% had this weird way of like learning as you<00:03:45.840> went<00:03:46.159> where<00:03:46.319> it<00:03:46.480> would<00:03:46.640> adapt<00:03:47.040> to<00:03:47.360> how<00:03:47.599> you 00:03:47.830 --> 00:03:47.840 align:start position:0% you went where it would adapt to how you 00:03:47.840 --> 00:03:49.990 align:start position:0% you went where it would adapt to how you were<00:03:48.000> coding<00:03:48.560> and<00:03:48.799> it<00:03:48.959> was<00:03:49.120> fast<00:03:49.280> as<00:03:49.519> hell.<00:03:49.760> It 00:03:49.990 --> 00:03:50.000 align:start position:0% were coding and it was fast as hell. It 00:03:50.000 --> 00:03:51.589 align:start position:0% were coding and it was fast as hell. It blew<00:03:50.159> me<00:03:50.319> away.<00:03:50.560> It<00:03:50.640> ended<00:03:50.879> up<00:03:50.959> being<00:03:51.120> my<00:03:51.360> first 00:03:51.589 --> 00:03:51.599 align:start position:0% blew me away. It ended up being my first 00:03:51.599 --> 00:03:53.190 align:start position:0% blew me away. It ended up being my first personal<00:03:52.080> investment<00:03:52.480> I<00:03:52.640> ever<00:03:52.799> made<00:03:52.959> where<00:03:53.120> I 00:03:53.190 --> 00:03:53.200 align:start position:0% personal investment I ever made where I 00:03:53.200 --> 00:03:54.789 align:start position:0% personal investment I ever made where I wired<00:03:53.519> my<00:03:53.680> own<00:03:53.840> money<00:03:54.080> instead<00:03:54.319> of<00:03:54.400> a<00:03:54.560> scout 00:03:54.789 --> 00:03:54.799 align:start position:0% wired my own money instead of a scout 00:03:54.799 --> 00:03:56.309 align:start position:0% wired my own money instead of a scout check<00:03:55.120> because<00:03:55.360> I<00:03:55.519> was<00:03:55.680> so<00:03:55.840> confident<00:03:56.159> in 00:03:56.309 --> 00:03:56.319 align:start position:0% check because I was so confident in 00:03:56.319 --> 00:03:58.149 align:start position:0% check because I was so confident in Super<00:03:56.640> Maven<00:03:56.959> and<00:03:57.120> specifically<00:03:57.519> in<00:03:57.760> Jacob. 00:03:58.149 --> 00:03:58.159 align:start position:0% Super Maven and specifically in Jacob. 00:03:58.159 --> 00:04:00.149 align:start position:0% Super Maven and specifically in Jacob. the<00:03:58.480> founder<00:03:58.720> of<00:03:58.879> Super<00:03:59.200> Maven<00:03:59.599> because<00:03:59.920> he 00:04:00.149 --> 00:04:00.159 align:start position:0% the founder of Super Maven because he 00:04:00.159 --> 00:04:02.630 align:start position:0% the founder of Super Maven because he came<00:04:00.400> from<00:04:00.640> Tab<00:04:00.959> 9.<00:04:01.280> Tab<00:04:01.680> 9<00:04:01.920> was<00:04:02.080> the<00:04:02.400> first 00:04:02.630 --> 00:04:02.640 align:start position:0% came from Tab 9. Tab 9 was the first 00:04:02.640 --> 00:04:04.869 align:start position:0% came from Tab 9. Tab 9 was the first attempt<00:04:03.040> at<00:04:03.200> an<00:04:03.439> AI<00:04:03.840> coding<00:04:04.159> thing<00:04:04.480> and<00:04:04.640> this 00:04:04.869 --> 00:04:04.879 align:start position:0% attempt at an AI coding thing and this 00:04:04.879 --> 00:04:06.949 align:start position:0% attempt at an AI coding thing and this started<00:04:05.280> all<00:04:05.439> the<00:04:05.680> way<00:04:05.760> back<00:04:06.000> in<00:04:06.319> what<00:04:06.640> yeah 00:04:06.949 --> 00:04:06.959 align:start position:0% started all the way back in what yeah 00:04:06.959 --> 00:04:10.630 align:start position:0% started all the way back in what yeah 2018<00:04:07.760> by<00:04:08.239> Jacob<00:04:08.640> Jackson.<00:04:09.280> I<00:04:09.519> was<00:04:10.000> blown<00:04:10.400> away 00:04:10.630 --> 00:04:10.640 align:start position:0% 2018 by Jacob Jackson. I was blown away 00:04:10.640 --> 00:04:12.869 align:start position:0% 2018 by Jacob Jackson. I was blown away to<00:04:10.799> see<00:04:10.959> him<00:04:11.200> leave<00:04:11.439> to<00:04:11.680> go<00:04:11.840> do<00:04:12.080> his<00:04:12.319> own<00:04:12.560> new 00:04:12.869 --> 00:04:12.879 align:start position:0% to see him leave to go do his own new 00:04:12.879 --> 00:04:14.949 align:start position:0% to see him leave to go do his own new thing<00:04:13.280> and<00:04:13.519> he<00:04:13.680> more<00:04:13.920> than<00:04:14.159> succeeded.<00:04:14.720> Super 00:04:14.949 --> 00:04:14.959 align:start position:0% thing and he more than succeeded. Super 00:04:14.959 --> 00:04:16.229 align:start position:0% thing and he more than succeeded. Super Maven<00:04:15.280> was<00:04:15.439> incredible.<00:04:15.760> There's<00:04:15.920> still<00:04:16.079> a 00:04:16.229 --> 00:04:16.239 align:start position:0% Maven was incredible. There's still a 00:04:16.239 --> 00:04:17.990 align:start position:0% Maven was incredible. There's still a bunch<00:04:16.400> of<00:04:16.479> people<00:04:16.639> who<00:04:16.880> dearly<00:04:17.359> miss<00:04:17.600> it.<00:04:17.840> So 00:04:17.990 --> 00:04:18.000 align:start position:0% bunch of people who dearly miss it. So 00:04:18.000 --> 00:04:19.590 align:start position:0% bunch of people who dearly miss it. So why<00:04:18.160> are<00:04:18.239> we<00:04:18.400> talking<00:04:18.560> about<00:04:18.720> Super<00:04:19.040> Maven? 00:04:19.590 --> 00:04:19.600 align:start position:0% why are we talking about Super Maven? 00:04:19.600 --> 00:04:21.430 align:start position:0% why are we talking about Super Maven? Well,<00:04:20.160> spoiler<00:04:20.479> if<00:04:20.720> you<00:04:20.799> weren't<00:04:20.959> here<00:04:21.120> at<00:04:21.359> the 00:04:21.430 --> 00:04:21.440 align:start position:0% Well, spoiler if you weren't here at the 00:04:21.440 --> 00:04:23.270 align:start position:0% Well, spoiler if you weren't here at the time.<00:04:21.919> Super<00:04:22.160> Maven<00:04:22.479> got<00:04:22.639> acquired<00:04:23.040> by 00:04:23.270 --> 00:04:23.280 align:start position:0% time. Super Maven got acquired by 00:04:23.280 --> 00:04:24.950 align:start position:0% time. Super Maven got acquired by Curser.<00:04:23.680> I<00:04:23.919> have<00:04:24.000> a<00:04:24.160> whole<00:04:24.320> video<00:04:24.560> about<00:04:24.720> that 00:04:24.950 --> 00:04:24.960 align:start position:0% Curser. I have a whole video about that 00:04:24.960 --> 00:04:26.310 align:start position:0% Curser. I have a whole video about that as<00:04:25.120> well<00:04:25.199> as<00:04:25.360> an<00:04:25.520> interview<00:04:25.759> I<00:04:25.919> did<00:04:26.000> with<00:04:26.160> both 00:04:26.310 --> 00:04:26.320 align:start position:0% as well as an interview I did with both 00:04:26.320 --> 00:04:28.230 align:start position:0% as well as an interview I did with both the<00:04:26.479> founders<00:04:26.720> of<00:04:26.880> Cursor<00:04:27.360> and<00:04:27.600> Super<00:04:27.840> Maven 00:04:28.230 --> 00:04:28.240 align:start position:0% the founders of Cursor and Super Maven 00:04:28.240 --> 00:04:30.070 align:start position:0% the founders of Cursor and Super Maven in<00:04:28.479> the<00:04:28.560> Cursor<00:04:29.040> office<00:04:29.360> back<00:04:29.520> when<00:04:29.680> it<00:04:29.840> was<00:04:29.919> a 00:04:30.070 --> 00:04:30.080 align:start position:0% in the Cursor office back when it was a 00:04:30.080 --> 00:04:32.710 align:start position:0% in the Cursor office back when it was a tiny<00:04:30.320> building<00:04:30.560> with<00:04:30.800> like<00:04:31.120> six<00:04:31.360> or<00:04:31.840> maybe<00:04:32.320> 10 00:04:32.710 --> 00:04:32.720 align:start position:0% tiny building with like six or maybe 10 00:04:32.720 --> 00:04:35.510 align:start position:0% tiny building with like six or maybe 10 people<00:04:33.040> max<00:04:33.440> in<00:04:33.600> it.<00:04:34.080> Great<00:04:34.320> time.<00:04:34.960> Wild<00:04:35.360> how 00:04:35.510 --> 00:04:35.520 align:start position:0% people max in it. Great time. Wild how 00:04:35.520 --> 00:04:36.710 align:start position:0% people max in it. Great time. Wild how much<00:04:35.680> things<00:04:35.840> have<00:04:36.000> changed.<00:04:36.240> They<00:04:36.479> have<00:04:36.560> like 00:04:36.710 --> 00:04:36.720 align:start position:0% much things have changed. They have like 00:04:36.720 --> 00:04:39.030 align:start position:0% much things have changed. They have like five<00:04:37.040> buildings<00:04:37.360> now.<00:04:37.759> But<00:04:37.919> Super<00:04:38.240> Maven<00:04:38.880> was 00:04:39.030 --> 00:04:39.040 align:start position:0% five buildings now. But Super Maven was 00:04:39.040 --> 00:04:40.870 align:start position:0% five buildings now. But Super Maven was a<00:04:39.280> great<00:04:39.520> project<00:04:39.840> and<00:04:40.080> it<00:04:40.240> really<00:04:40.479> showed 00:04:40.870 --> 00:04:40.880 align:start position:0% a great project and it really showed 00:04:40.880 --> 00:04:42.950 align:start position:0% a great project and it really showed what<00:04:41.199> can<00:04:41.360> happen<00:04:41.759> if<00:04:42.080> you<00:04:42.320> try<00:04:42.479> to<00:04:42.720> do 00:04:42.950 --> 00:04:42.960 align:start position:0% what can happen if you try to do 00:04:42.960 --> 00:04:45.110 align:start position:0% what can happen if you try to do everything<00:04:43.360> from<00:04:43.680> the<00:04:43.840> user<00:04:44.160> experience<00:04:44.880> to 00:04:45.110 --> 00:04:45.120 align:start position:0% everything from the user experience to 00:04:45.120 --> 00:04:48.070 align:start position:0% everything from the user experience to the<00:04:45.440> model<00:04:45.759> itself<00:04:46.479> around<00:04:47.040> specific<00:04:47.600> focused 00:04:48.070 --> 00:04:48.080 align:start position:0% the model itself around specific focused 00:04:48.080 --> 00:04:50.469 align:start position:0% the model itself around specific focused coding<00:04:48.560> tasks.<00:04:49.280> And<00:04:49.440> it's<00:04:49.680> clear,<00:04:50.080> if<00:04:50.320> you 00:04:50.469 --> 00:04:50.479 align:start position:0% coding tasks. And it's clear, if you 00:04:50.479 --> 00:04:52.390 align:start position:0% coding tasks. And it's clear, if you couldn't<00:04:50.880> tell<00:04:51.040> by<00:04:51.280> now,<00:04:51.520> that<00:04:51.759> Jacob,<00:04:52.240> the 00:04:52.390 --> 00:04:52.400 align:start position:0% couldn't tell by now, that Jacob, the 00:04:52.400 --> 00:04:55.270 align:start position:0% couldn't tell by now, that Jacob, the founder<00:04:52.560> of<00:04:52.720> Super<00:04:52.960> Maven,<00:04:53.520> is<00:04:54.080> obsessed<00:04:54.960> with 00:04:55.270 --> 00:04:55.280 align:start position:0% founder of Super Maven, is obsessed with 00:04:55.280 --> 00:04:58.070 align:start position:0% founder of Super Maven, is obsessed with making<00:04:55.680> AI<00:04:56.240> really,<00:04:56.560> really<00:04:56.960> good<00:04:57.280> at<00:04:57.680> code 00:04:58.070 --> 00:04:58.080 align:start position:0% making AI really, really good at code 00:04:58.080 --> 00:05:00.070 align:start position:0% making AI really, really good at code and<00:04:58.479> nothing<00:04:58.800> else.<00:04:59.199> When<00:04:59.360> he<00:04:59.520> joined<00:04:59.680> Cursor, 00:05:00.070 --> 00:05:00.080 align:start position:0% and nothing else. When he joined Cursor, 00:05:00.080 --> 00:05:01.670 align:start position:0% and nothing else. When he joined Cursor, him<00:05:00.160> and<00:05:00.320> his<00:05:00.479> team<00:05:00.720> went<00:05:00.880> allin<00:05:01.360> on 00:05:01.670 --> 00:05:01.680 align:start position:0% him and his team went allin on 00:05:01.680 --> 00:05:03.270 align:start position:0% him and his team went allin on autocomplete,<00:05:02.320> which<00:05:02.479> is<00:05:02.560> why<00:05:02.720> to<00:05:02.960> this<00:05:03.040> day, 00:05:03.270 --> 00:05:03.280 align:start position:0% autocomplete, which is why to this day, 00:05:03.280 --> 00:05:04.950 align:start position:0% autocomplete, which is why to this day, Cursor's<00:05:03.680> autocomplete<00:05:04.240> is<00:05:04.400> just<00:05:04.560> so<00:05:04.800> much 00:05:04.950 --> 00:05:04.960 align:start position:0% Cursor's autocomplete is just so much 00:05:04.960 --> 00:05:06.390 align:start position:0% Cursor's autocomplete is just so much better<00:05:05.199> than<00:05:05.440> anything<00:05:05.759> else.<00:05:06.160> It's<00:05:06.240> the 00:05:06.390 --> 00:05:06.400 align:start position:0% better than anything else. It's the 00:05:06.400 --> 00:05:07.670 align:start position:0% better than anything else. It's the thing<00:05:06.479> that<00:05:06.639> keeps<00:05:06.800> me<00:05:06.960> going<00:05:07.120> back.<00:05:07.440> I've 00:05:07.670 --> 00:05:07.680 align:start position:0% thing that keeps me going back. I've 00:05:07.680 --> 00:05:09.430 align:start position:0% thing that keeps me going back. I've been<00:05:07.840> trying<00:05:07.919> to<00:05:08.080> move<00:05:08.240> to<00:05:08.400> VS<00:05:08.720> Code,<00:05:08.880> cuz<00:05:09.199> to 00:05:09.430 --> 00:05:09.440 align:start position:0% been trying to move to VS Code, cuz to 00:05:09.440 --> 00:05:11.670 align:start position:0% been trying to move to VS Code, cuz to be<00:05:09.680> frank,<00:05:10.080> it's<00:05:10.240> a<00:05:10.479> much<00:05:10.639> lighter<00:05:11.199> thing. 00:05:11.670 --> 00:05:11.680 align:start position:0% be frank, it's a much lighter thing. 00:05:11.680 --> 00:05:13.270 align:start position:0% be frank, it's a much lighter thing. It's<00:05:11.919> less<00:05:12.160> likely<00:05:12.400> to<00:05:12.560> cause<00:05:12.800> me<00:05:12.960> problems 00:05:13.270 --> 00:05:13.280 align:start position:0% It's less likely to cause me problems 00:05:13.280 --> 00:05:15.110 align:start position:0% It's less likely to cause me problems when<00:05:13.520> I<00:05:13.600> use<00:05:13.759> it.<00:05:14.080> I've<00:05:14.320> been<00:05:14.479> very<00:05:14.639> happy<00:05:14.880> with 00:05:15.110 --> 00:05:15.120 align:start position:0% when I use it. I've been very happy with 00:05:15.120 --> 00:05:17.110 align:start position:0% when I use it. I've been very happy with VS<00:05:15.520> Code<00:05:15.759> just<00:05:16.080> not<00:05:16.320> crashing<00:05:16.720> and<00:05:16.880> being 00:05:17.110 --> 00:05:17.120 align:start position:0% VS Code just not crashing and being 00:05:17.120 --> 00:05:18.629 align:start position:0% VS Code just not crashing and being fast,<00:05:17.520> but<00:05:17.680> as<00:05:17.840> soon<00:05:18.000> as<00:05:18.080> I<00:05:18.160> have<00:05:18.320> to<00:05:18.400> actually 00:05:18.629 --> 00:05:18.639 align:start position:0% fast, but as soon as I have to actually 00:05:18.639 --> 00:05:20.150 align:start position:0% fast, but as soon as I have to actually type<00:05:18.800> in<00:05:18.960> it,<00:05:19.199> I<00:05:19.360> find<00:05:19.520> myself<00:05:19.759> sitting<00:05:20.000> there 00:05:20.150 --> 00:05:20.160 align:start position:0% type in it, I find myself sitting there 00:05:20.160 --> 00:05:21.510 align:start position:0% type in it, I find myself sitting there waiting<00:05:20.400> for<00:05:20.479> the<00:05:20.639> autocomplete.<00:05:21.199> And<00:05:21.280> when 00:05:21.510 --> 00:05:21.520 align:start position:0% waiting for the autocomplete. And when 00:05:21.520 --> 00:05:23.350 align:start position:0% waiting for the autocomplete. And when eventually<00:05:22.000> Copilot<00:05:22.560> does<00:05:22.720> kick<00:05:22.960> in,<00:05:23.199> it 00:05:23.350 --> 00:05:23.360 align:start position:0% eventually Copilot does kick in, it 00:05:23.360 --> 00:05:25.110 align:start position:0% eventually Copilot does kick in, it gives<00:05:23.520> a<00:05:23.919> recommendation<00:05:24.479> and<00:05:24.720> I'm<00:05:24.880> just 00:05:25.110 --> 00:05:25.120 align:start position:0% gives a recommendation and I'm just 00:05:25.120 --> 00:05:27.430 align:start position:0% gives a recommendation and I'm just so<00:05:25.360> used<00:05:25.600> to<00:05:25.840> using<00:05:26.160> cursors<00:05:26.880> because<00:05:27.120> of<00:05:27.280> the 00:05:27.430 --> 00:05:27.440 align:start position:0% so used to using cursors because of the 00:05:27.440 --> 00:05:29.510 align:start position:0% so used to using cursors because of the the<00:05:27.759> Super<00:05:28.000> Maven<00:05:28.479> magic<00:05:28.800> that<00:05:29.120> Jacob 00:05:29.510 --> 00:05:29.520 align:start position:0% the Super Maven magic that Jacob 00:05:29.520 --> 00:05:31.430 align:start position:0% the Super Maven magic that Jacob injected<00:05:29.919> into<00:05:30.080> it.<00:05:30.320> But<00:05:30.479> Jacob<00:05:30.880> got<00:05:31.039> tired<00:05:31.280> of 00:05:31.430 --> 00:05:31.440 align:start position:0% injected into it. But Jacob got tired of 00:05:31.440 --> 00:05:33.110 align:start position:0% injected into it. But Jacob got tired of just<00:05:31.680> working<00:05:31.919> on<00:05:32.160> autocomplete.<00:05:32.880> Obviously, 00:05:33.110 --> 00:05:33.120 align:start position:0% just working on autocomplete. Obviously, 00:05:33.120 --> 00:05:34.390 align:start position:0% just working on autocomplete. Obviously, his<00:05:33.280> tag<00:05:33.520> complete<00:05:33.759> was<00:05:33.840> great<00:05:34.000> and<00:05:34.160> he<00:05:34.240> had<00:05:34.320> to 00:05:34.390 --> 00:05:34.400 align:start position:0% his tag complete was great and he had to 00:05:34.400 --> 00:05:35.590 align:start position:0% his tag complete was great and he had to build<00:05:34.560> custom<00:05:34.880> models<00:05:35.039> and<00:05:35.280> all<00:05:35.360> of<00:05:35.440> these 00:05:35.590 --> 00:05:35.600 align:start position:0% build custom models and all of these 00:05:35.600 --> 00:05:37.110 align:start position:0% build custom models and all of these things<00:05:35.680> for<00:05:35.919> it<00:05:36.080> internally,<00:05:36.639> but<00:05:36.800> he<00:05:36.960> wanted 00:05:37.110 --> 00:05:37.120 align:start position:0% things for it internally, but he wanted 00:05:37.120 --> 00:05:38.629 align:start position:0% things for it internally, but he wanted to<00:05:37.280> go<00:05:37.360> further.<00:05:37.840> That's<00:05:38.080> why<00:05:38.240> he<00:05:38.400> made 00:05:38.629 --> 00:05:38.639 align:start position:0% to go further. That's why he made 00:05:38.639 --> 00:05:40.950 align:start position:0% to go further. That's why he made composer.<00:05:39.520> Composer<00:05:40.000> one<00:05:40.240> was<00:05:40.400> Cursor's 00:05:40.950 --> 00:05:40.960 align:start position:0% composer. Composer one was Cursor's 00:05:40.960 --> 00:05:43.670 align:start position:0% composer. Composer one was Cursor's first<00:05:41.199> attempt<00:05:41.520> at<00:05:41.759> making<00:05:42.000> a<00:05:42.240> new<00:05:42.800> model<00:05:43.280> for 00:05:43.670 --> 00:05:43.680 align:start position:0% first attempt at making a new model for 00:05:43.680 --> 00:05:45.990 align:start position:0% first attempt at making a new model for full<00:05:44.080> code<00:05:44.400> generation<00:05:44.880> from<00:05:45.199> scratch.<00:05:45.680> And<00:05:45.840> I 00:05:45.990 --> 00:05:46.000 align:start position:0% full code generation from scratch. And I 00:05:46.000 --> 00:05:47.270 align:start position:0% full code generation from scratch. And I could<00:05:46.080> see<00:05:46.160> why<00:05:46.400> Jacob<00:05:46.720> would<00:05:46.960> want<00:05:47.120> to<00:05:47.199> do 00:05:47.270 --> 00:05:47.280 align:start position:0% could see why Jacob would want to do 00:05:47.280 --> 00:05:48.790 align:start position:0% could see why Jacob would want to do this.<00:05:47.680> When<00:05:47.840> you<00:05:48.000> look<00:05:48.080> at<00:05:48.160> his<00:05:48.400> history, 00:05:48.790 --> 00:05:48.800 align:start position:0% this. When you look at his history, 00:05:48.800 --> 00:05:50.230 align:start position:0% this. When you look at his history, building<00:05:49.120> models<00:05:49.360> for<00:05:49.520> all<00:05:49.680> sorts<00:05:49.919> of<00:05:50.000> code 00:05:50.230 --> 00:05:50.240 align:start position:0% building models for all sorts of code 00:05:50.240 --> 00:05:52.390 align:start position:0% building models for all sorts of code tasks,<00:05:50.880> as<00:05:51.039> well<00:05:51.199> as<00:05:51.360> the<00:05:51.680> amount<00:05:51.919> of<00:05:52.080> data 00:05:52.390 --> 00:05:52.400 align:start position:0% tasks, as well as the amount of data 00:05:52.400 --> 00:05:55.029 align:start position:0% tasks, as well as the amount of data that<00:05:52.720> cursor<00:05:53.199> has<00:05:53.520> on<00:05:53.919> chat<00:05:54.400> histories<00:05:54.800> and 00:05:55.029 --> 00:05:55.039 align:start position:0% that cursor has on chat histories and 00:05:55.039 --> 00:05:56.390 align:start position:0% that cursor has on chat histories and people<00:05:55.280> building<00:05:55.520> in<00:05:55.680> these<00:05:55.919> projects, 00:05:56.390 --> 00:05:56.400 align:start position:0% people building in these projects, 00:05:56.400 --> 00:05:57.670 align:start position:0% people building in these projects, you'll<00:05:56.639> realize<00:05:56.880> that<00:05:56.960> he's<00:05:57.280> kind<00:05:57.360> of<00:05:57.440> set<00:05:57.600> up 00:05:57.670 --> 00:05:57.680 align:start position:0% you'll realize that he's kind of set up 00:05:57.680 --> 00:05:59.110 align:start position:0% you'll realize that he's kind of set up for<00:05:57.919> success<00:05:58.160> here.<00:05:58.400> He<00:05:58.560> has<00:05:58.639> all<00:05:58.800> the<00:05:58.960> things 00:05:59.110 --> 00:05:59.120 align:start position:0% for success here. He has all the things 00:05:59.120 --> 00:06:01.189 align:start position:0% for success here. He has all the things he<00:05:59.280> needs<00:05:59.440> to<00:05:59.680> do<00:05:59.840> something<00:06:00.320> awesome.<00:06:00.960> But<00:06:01.039> it 00:06:01.189 --> 00:06:01.199 align:start position:0% he needs to do something awesome. But it 00:06:01.199 --> 00:06:03.189 align:start position:0% he needs to do something awesome. But it also<00:06:01.440> takes<00:06:01.680> a<00:06:01.840> ton<00:06:02.000> of<00:06:02.160> money<00:06:02.479> and<00:06:02.800> research 00:06:03.189 --> 00:06:03.199 align:start position:0% also takes a ton of money and research 00:06:03.199 --> 00:06:05.350 align:start position:0% also takes a ton of money and research and<00:06:03.440> people<00:06:03.759> and<00:06:04.000> GPUs<00:06:04.639> to<00:06:04.880> do<00:06:05.039> something 00:06:05.350 --> 00:06:05.360 align:start position:0% and people and GPUs to do something 00:06:05.360 --> 00:06:07.189 align:start position:0% and people and GPUs to do something crazy.<00:06:05.840> He<00:06:06.080> still<00:06:06.240> managed<00:06:06.479> to<00:06:06.639> with<00:06:06.800> Composer 00:06:07.189 --> 00:06:07.199 align:start position:0% crazy. He still managed to with Composer 00:06:07.199 --> 00:06:08.950 align:start position:0% crazy. He still managed to with Composer 1,<00:06:07.440> though.<00:06:07.600> Like<00:06:07.759> that<00:06:07.919> model<00:06:08.240> blew<00:06:08.479> me<00:06:08.639> away. 00:06:08.950 --> 00:06:08.960 align:start position:0% 1, though. Like that model blew me away. 00:06:08.960 --> 00:06:10.790 align:start position:0% 1, though. Like that model blew me away. was<00:06:09.199> so<00:06:09.360> much<00:06:09.600> faster<00:06:09.840> than<00:06:10.080> everything<00:06:10.400> else 00:06:10.790 --> 00:06:10.800 align:start position:0% was so much faster than everything else 00:06:10.800 --> 00:06:13.270 align:start position:0% was so much faster than everything else and<00:06:11.120> just<00:06:11.440> it<00:06:11.680> was<00:06:11.919> good.<00:06:12.319> It<00:06:12.560> didn't<00:06:12.960> wasn't 00:06:13.270 --> 00:06:13.280 align:start position:0% and just it was good. It didn't wasn't 00:06:13.280 --> 00:06:14.870 align:start position:0% and just it was good. It didn't wasn't great.<00:06:13.600> It<00:06:13.840> didn't<00:06:14.000> have<00:06:14.240> the<00:06:14.400> like<00:06:14.560> deep 00:06:14.870 --> 00:06:14.880 align:start position:0% great. It didn't have the like deep 00:06:14.880 --> 00:06:15.909 align:start position:0% great. It didn't have the like deep intelligence<00:06:15.280> and<00:06:15.440> understanding<00:06:15.759> of<00:06:15.840> my 00:06:15.909 --> 00:06:15.919 align:start position:0% intelligence and understanding of my 00:06:15.919 --> 00:06:17.430 align:start position:0% intelligence and understanding of my code<00:06:16.080> bases<00:06:16.400> that<00:06:16.560> other<00:06:16.720> models<00:06:17.039> did<00:06:17.120> at<00:06:17.280> the 00:06:17.430 --> 00:06:17.440 align:start position:0% code bases that other models did at the 00:06:17.440 --> 00:06:19.270 align:start position:0% code bases that other models did at the time,<00:06:17.759> but<00:06:18.000> god<00:06:18.240> damn<00:06:18.479> was<00:06:18.639> it<00:06:18.800> capable<00:06:19.120> of 00:06:19.270 --> 00:06:19.280 align:start position:0% time, but god damn was it capable of 00:06:19.280 --> 00:06:20.550 align:start position:0% time, but god damn was it capable of doing<00:06:19.440> incredible<00:06:19.840> stuff.<00:06:20.080> And<00:06:20.240> that's<00:06:20.400> why 00:06:20.550 --> 00:06:20.560 align:start position:0% doing incredible stuff. And that's why 00:06:20.560 --> 00:06:21.990 align:start position:0% doing incredible stuff. And that's why they<00:06:20.720> kept<00:06:20.880> on<00:06:21.039> iterating.<00:06:21.520> That's<00:06:21.680> why<00:06:21.840> he 00:06:21.990 --> 00:06:22.000 align:start position:0% they kept on iterating. That's why he 00:06:22.000 --> 00:06:23.270 align:start position:0% they kept on iterating. That's why he got<00:06:22.080> to<00:06:22.240> hire<00:06:22.479> more.<00:06:22.639> That's<00:06:22.800> why<00:06:22.960> they<00:06:23.120> put 00:06:23.270 --> 00:06:23.280 align:start position:0% got to hire more. That's why they put 00:06:23.280 --> 00:06:24.950 align:start position:0% got to hire more. That's why they put more<00:06:23.440> money<00:06:23.600> into<00:06:23.840> GPUs,<00:06:24.400> more<00:06:24.560> effort<00:06:24.720> into 00:06:24.950 --> 00:06:24.960 align:start position:0% more money into GPUs, more effort into 00:06:24.960 --> 00:06:26.390 align:start position:0% more money into GPUs, more effort into curating<00:06:25.280> all<00:06:25.440> their<00:06:25.600> data,<00:06:25.840> and<00:06:26.000> just<00:06:26.240> did 00:06:26.390 --> 00:06:26.400 align:start position:0% curating all their data, and just did 00:06:26.400 --> 00:06:28.070 align:start position:0% curating all their data, and just did everything<00:06:26.720> they<00:06:27.039> had<00:06:27.120> to<00:06:27.280> to<00:06:27.520> set<00:06:27.680> themselves 00:06:28.070 --> 00:06:28.080 align:start position:0% everything they had to to set themselves 00:06:28.080 --> 00:06:29.909 align:start position:0% everything they had to to set themselves up<00:06:28.479> to<00:06:28.720> make<00:06:28.880> better<00:06:29.199> models.<00:06:29.600> But<00:06:29.759> now 00:06:29.909 --> 00:06:29.919 align:start position:0% up to make better models. But now 00:06:29.919 --> 00:06:31.990 align:start position:0% up to make better models. But now there's<00:06:30.160> the<00:06:30.400> question<00:06:30.560> of<00:06:31.120> why<00:06:31.680> making 00:06:31.990 --> 00:06:32.000 align:start position:0% there's the question of why making 00:06:32.000 --> 00:06:33.830 align:start position:0% there's the question of why making models<00:06:32.319> is<00:06:32.560> super<00:06:32.880> expensive<00:06:33.440> and<00:06:33.600> the 00:06:33.830 --> 00:06:33.840 align:start position:0% models is super expensive and the 00:06:33.840 --> 00:06:35.110 align:start position:0% models is super expensive and the chances<00:06:34.000> that<00:06:34.160> you<00:06:34.319> can<00:06:34.400> make<00:06:34.479> a<00:06:34.639> model<00:06:34.880> as 00:06:35.110 --> 00:06:35.120 align:start position:0% chances that you can make a model as 00:06:35.120 --> 00:06:37.510 align:start position:0% chances that you can make a model as good<00:06:35.280> as<00:06:35.520> Frontier<00:06:36.000> Labs<00:06:36.400> at<00:06:36.720> anything<00:06:37.280> is 00:06:37.510 --> 00:06:37.520 align:start position:0% good as Frontier Labs at anything is 00:06:37.520 --> 00:06:39.510 align:start position:0% good as Frontier Labs at anything is very<00:06:37.759> low<00:06:38.000> at<00:06:38.160> best.<00:06:38.560> Why<00:06:38.800> would<00:06:38.960> they<00:06:39.120> put<00:06:39.280> all 00:06:39.510 --> 00:06:39.520 align:start position:0% very low at best. Why would they put all 00:06:39.520 --> 00:06:41.270 align:start position:0% very low at best. Why would they put all of<00:06:39.600> this<00:06:39.840> time<00:06:40.160> into<00:06:40.479> building<00:06:40.800> something<00:06:41.039> of 00:06:41.270 --> 00:06:41.280 align:start position:0% of this time into building something of 00:06:41.280 --> 00:06:43.670 align:start position:0% of this time into building something of their<00:06:41.440> own<00:06:41.600> like<00:06:41.840> this?<00:06:42.560> Well,<00:06:43.280> they<00:06:43.520> have 00:06:43.670 --> 00:06:43.680 align:start position:0% their own like this? Well, they have 00:06:43.680 --> 00:06:46.469 align:start position:0% their own like this? Well, they have some<00:06:43.919> competition.<00:06:44.960> I<00:06:45.199> want<00:06:45.280> to<00:06:45.440> ask<00:06:45.759> y'all<00:06:46.160> a 00:06:46.469 --> 00:06:46.479 align:start position:0% some competition. I want to ask y'all a 00:06:46.479 --> 00:06:48.469 align:start position:0% some competition. I want to ask y'all a question,<00:06:46.880> chat.<00:06:47.280> And<00:06:47.440> do<00:06:47.680> not<00:06:47.919> answer<00:06:48.240> this 00:06:48.469 --> 00:06:48.479 align:start position:0% question, chat. And do not answer this 00:06:48.479 --> 00:06:50.629 align:start position:0% question, chat. And do not answer this question<00:06:48.800> if<00:06:49.039> you<00:06:49.280> already<00:06:49.600> know<00:06:49.919> the<00:06:50.240> answer. 00:06:50.629 --> 00:06:50.639 align:start position:0% question if you already know the answer. 00:06:50.639 --> 00:06:52.629 align:start position:0% question if you already know the answer. To<00:06:50.800> everybody<00:06:51.120> who<00:06:51.440> doesn't,<00:06:51.919> I<00:06:52.160> want<00:06:52.319> you<00:06:52.479> to 00:06:52.629 --> 00:06:52.639 align:start position:0% To everybody who doesn't, I want you to 00:06:52.639 --> 00:06:55.510 align:start position:0% To everybody who doesn't, I want you to guess.<00:06:53.520> How<00:06:53.759> much<00:06:54.080> inference<00:06:54.880> do<00:06:55.039> you<00:06:55.280> think 00:06:55.510 --> 00:06:55.520 align:start position:0% guess. How much inference do you think 00:06:55.520 --> 00:06:58.710 align:start position:0% guess. How much inference do you think you<00:06:55.759> can<00:06:55.919> do<00:06:56.160> per<00:06:56.479> month<00:06:56.800> on<00:06:57.039> the<00:06:57.360> $200<00:06:58.160> a<00:06:58.400> month 00:06:58.710 --> 00:06:58.720 align:start position:0% you can do per month on the $200 a month 00:06:58.720 --> 00:07:00.469 align:start position:0% you can do per month on the $200 a month Claude<00:06:59.120> Code<00:06:59.440> plan?<00:06:59.919> Like<00:07:00.080> if<00:07:00.240> you<00:07:00.400> were 00:07:00.469 --> 00:07:00.479 align:start position:0% Claude Code plan? Like if you were 00:07:00.479 --> 00:07:02.870 align:start position:0% Claude Code plan? Like if you were paying<00:07:00.800> full<00:07:01.120> API<00:07:01.599> rates<00:07:02.160> and<00:07:02.400> you<00:07:02.560> max<00:07:02.800> out 00:07:02.870 --> 00:07:02.880 align:start position:0% paying full API rates and you max out 00:07:02.880 --> 00:07:04.870 align:start position:0% paying full API rates and you max out your<00:07:03.120> $200<00:07:03.520> a<00:07:03.680> month<00:07:03.840> plan<00:07:04.000> on<00:07:04.160> Cloud<00:07:04.479> Code, 00:07:04.870 --> 00:07:04.880 align:start position:0% your $200 a month plan on Cloud Code, 00:07:04.880 --> 00:07:07.029 align:start position:0% your $200 a month plan on Cloud Code, how<00:07:05.039> much<00:07:05.199> money<00:07:05.599> of<00:07:05.840> API<00:07:06.319> usage<00:07:06.639> do<00:07:06.800> you<00:07:06.960> think 00:07:07.029 --> 00:07:07.039 align:start position:0% how much money of API usage do you think 00:07:07.039 --> 00:07:09.110 align:start position:0% how much money of API usage do you think you<00:07:07.280> can<00:07:07.360> do?<00:07:07.680> Just<00:07:07.919> drop<00:07:08.240> numbers<00:07:08.560> in<00:07:08.720> chat. 00:07:09.110 --> 00:07:09.120 align:start position:0% you can do? Just drop numbers in chat. 00:07:09.120 --> 00:07:10.469 align:start position:0% you can do? Just drop numbers in chat. You're<00:07:09.280> putting<00:07:09.440> in<00:07:09.599> 200<00:07:09.919> bucks.<00:07:10.240> What<00:07:10.319> do<00:07:10.400> you 00:07:10.469 --> 00:07:10.479 align:start position:0% You're putting in 200 bucks. What do you 00:07:10.479 --> 00:07:13.350 align:start position:0% You're putting in 200 bucks. What do you think<00:07:10.560> you<00:07:10.720> can<00:07:10.880> get<00:07:10.960> out?<00:07:11.280> Like<00:07:11.840> 200,<00:07:12.560> 500, 00:07:13.350 --> 00:07:13.360 align:start position:0% think you can get out? Like 200, 500, 00:07:13.360 --> 00:07:16.629 align:start position:0% think you can get out? Like 200, 500, 2,000?<00:07:14.400> Okay,<00:07:14.639> we<00:07:14.800> see<00:07:14.960> numbers<00:07:15.360> like<00:07:15.919> 2,000, 00:07:16.629 --> 00:07:16.639 align:start position:0% 2,000? Okay, we see numbers like 2,000, 00:07:16.639 --> 00:07:19.510 align:start position:0% 2,000? Okay, we see numbers like 2,000, 1,600,<00:07:17.599> 3,000,<00:07:18.240> 2K,<00:07:18.639> 1K.<00:07:19.199> Remember,<00:07:19.360> you're 00:07:19.510 --> 00:07:19.520 align:start position:0% 1,600, 3,000, 2K, 1K. Remember, you're 00:07:19.520 --> 00:07:20.950 align:start position:0% 1,600, 3,000, 2K, 1K. Remember, you're paying<00:07:19.759> 200<00:07:20.000> bucks.<00:07:20.319> So<00:07:20.479> getting<00:07:20.720> anything 00:07:20.950 --> 00:07:20.960 align:start position:0% paying 200 bucks. So getting anything 00:07:20.960 --> 00:07:23.029 align:start position:0% paying 200 bucks. So getting anything over<00:07:21.280> 200<00:07:21.840> is<00:07:22.080> a<00:07:22.240> deal.<00:07:22.800> you<00:07:22.960> could 00:07:23.029 --> 00:07:23.039 align:start position:0% over 200 is a deal. you could 00:07:23.039 --> 00:07:25.350 align:start position:0% over 200 is a deal. you could theoretically<00:07:23.520> get<00:07:23.599> way<00:07:23.840> more.<00:07:24.400> So,<00:07:25.199> here's 00:07:25.350 --> 00:07:25.360 align:start position:0% theoretically get way more. So, here's 00:07:25.360 --> 00:07:26.950 align:start position:0% theoretically get way more. So, here's where<00:07:25.520> things<00:07:25.759> get<00:07:25.919> interesting.<00:07:26.400> When<00:07:26.639> Cloud 00:07:26.950 --> 00:07:26.960 align:start position:0% where things get interesting. When Cloud 00:07:26.960 --> 00:07:29.110 align:start position:0% where things get interesting. When Cloud Code<00:07:27.280> first<00:07:27.599> introduced<00:07:28.080> the<00:07:28.319> $200<00:07:28.800> a<00:07:28.880> month 00:07:29.110 --> 00:07:29.120 align:start position:0% Code first introduced the $200 a month 00:07:29.120 --> 00:07:31.589 align:start position:0% Code first introduced the $200 a month sub,<00:07:29.599> the<00:07:29.840> answer<00:07:30.080> was<00:07:30.319> around<00:07:30.720> $1,500<00:07:31.440> to 00:07:31.589 --> 00:07:31.599 align:start position:0% sub, the answer was around $1,500 to 00:07:31.599 --> 00:07:33.990 align:start position:0% sub, the answer was around $1,500 to $2,000,<00:07:32.639> but<00:07:32.800> they<00:07:32.960> have<00:07:33.199> continued<00:07:33.599> ramping 00:07:33.990 --> 00:07:34.000 align:start position:0% $2,000, but they have continued ramping 00:07:34.000 --> 00:07:35.990 align:start position:0% $2,000, but they have continued ramping it<00:07:34.160> up<00:07:34.560> since.<00:07:35.199> According<00:07:35.440> to<00:07:35.520> Kurser's 00:07:35.990 --> 00:07:36.000 align:start position:0% it up since. According to Kurser's 00:07:36.000 --> 00:07:37.430 align:start position:0% it up since. According to Kurser's internal<00:07:36.319> analysis,<00:07:36.800> their<00:07:37.039> last<00:07:37.199> year 00:07:37.430 --> 00:07:37.440 align:start position:0% internal analysis, their last year 00:07:37.440 --> 00:07:38.950 align:start position:0% internal analysis, their last year estimate<00:07:37.680> was<00:07:37.840> that<00:07:38.080> $200<00:07:38.400> per<00:07:38.560> month<00:07:38.720> cloud 00:07:38.950 --> 00:07:38.960 align:start position:0% estimate was that $200 per month cloud 00:07:38.960 --> 00:07:41.029 align:start position:0% estimate was that $200 per month cloud code<00:07:39.120> sub<00:07:39.360> gave<00:07:39.440> you<00:07:39.680> $2,000<00:07:40.160> in<00:07:40.319> compute,<00:07:40.800> but 00:07:41.029 --> 00:07:41.039 align:start position:0% code sub gave you $2,000 in compute, but 00:07:41.039 --> 00:07:42.390 align:start position:0% code sub gave you $2,000 in compute, but now<00:07:41.199> the<00:07:41.360> subsidiation<00:07:41.919> appeared<00:07:42.160> more 00:07:42.390 --> 00:07:42.400 align:start position:0% now the subsidiation appeared more 00:07:42.400 --> 00:07:44.230 align:start position:0% now the subsidiation appeared more aggressive<00:07:42.800> with<00:07:43.120> $200<00:07:43.520> plans<00:07:43.840> able<00:07:44.080> to 00:07:44.230 --> 00:07:44.240 align:start position:0% aggressive with $200 plans able to 00:07:44.240 --> 00:07:47.029 align:start position:0% aggressive with $200 plans able to consume<00:07:44.560> about<00:07:45.120> $5,000<00:07:46.080> in<00:07:46.319> compute.<00:07:46.800> Do<00:07:46.960> you 00:07:47.029 --> 00:07:47.039 align:start position:0% consume about $5,000 in compute. Do you 00:07:47.039 --> 00:07:49.510 align:start position:0% consume about $5,000 in compute. Do you see<00:07:47.199> the<00:07:47.440> problem<00:07:47.680> here?<00:07:48.400> That<00:07:48.560> is<00:07:48.800> insane. 00:07:49.510 --> 00:07:49.520 align:start position:0% see the problem here? That is insane. 00:07:49.520 --> 00:07:52.070 align:start position:0% see the problem here? That is insane. This<00:07:49.759> level<00:07:50.000> of<00:07:50.240> subsidization<00:07:51.199> is<00:07:51.520> absurd. 00:07:52.070 --> 00:07:52.080 align:start position:0% This level of subsidization is absurd. 00:07:52.080 --> 00:07:53.909 align:start position:0% This level of subsidization is absurd. And<00:07:52.240> I<00:07:52.400> want<00:07:52.479> to<00:07:52.639> be<00:07:52.720> clear,<00:07:53.120> that's<00:07:53.440> not<00:07:53.680> going 00:07:53.909 --> 00:07:53.919 align:start position:0% And I want to be clear, that's not going 00:07:53.919 --> 00:07:55.589 align:start position:0% And I want to be clear, that's not going to<00:07:54.080> last.<00:07:54.560> Anthropic<00:07:55.120> is<00:07:55.199> not<00:07:55.360> going<00:07:55.440> to<00:07:55.520> be 00:07:55.589 --> 00:07:55.599 align:start position:0% to last. Anthropic is not going to be 00:07:55.599 --> 00:07:57.270 align:start position:0% to last. Anthropic is not going to be able<00:07:55.759> to<00:07:55.919> keep<00:07:56.160> giving<00:07:56.400> you<00:07:56.560> this<00:07:56.800> insane 00:07:57.270 --> 00:07:57.280 align:start position:0% able to keep giving you this insane 00:07:57.280 --> 00:07:58.950 align:start position:0% able to keep giving you this insane amount<00:07:57.440> of<00:07:57.599> inference<00:07:58.080> for<00:07:58.319> that<00:07:58.560> type<00:07:58.800> of 00:07:58.950 --> 00:07:58.960 align:start position:0% amount of inference for that type of 00:07:58.960 --> 00:08:01.270 align:start position:0% amount of inference for that type of price.<00:07:59.759> They<00:08:00.000> just<00:08:00.160> they<00:08:00.400> can't.<00:08:00.720> That's<00:08:01.039> not 00:08:01.270 --> 00:08:01.280 align:start position:0% price. They just they can't. That's not 00:08:01.280 --> 00:08:03.110 align:start position:0% price. They just they can't. That's not realistic.<00:08:01.919> They<00:08:02.080> just<00:08:02.240> want<00:08:02.319> to<00:08:02.560> price 00:08:03.110 --> 00:08:03.120 align:start position:0% realistic. They just want to price 00:08:03.120 --> 00:08:04.710 align:start position:0% realistic. They just want to price everyone<00:08:03.520> out.<00:08:04.000> That's<00:08:04.240> why<00:08:04.400> they're<00:08:04.560> doing 00:08:04.710 --> 00:08:04.720 align:start position:0% everyone out. That's why they're doing 00:08:04.720 --> 00:08:06.150 align:start position:0% everyone out. That's why they're doing this.<00:08:04.960> That's<00:08:05.120> why<00:08:05.280> they're<00:08:05.440> subsidizing<00:08:06.000> so 00:08:06.150 --> 00:08:06.160 align:start position:0% this. That's why they're subsidizing so 00:08:06.160 --> 00:08:07.830 align:start position:0% this. That's why they're subsidizing so hard.<00:08:06.479> And<00:08:06.639> for<00:08:06.800> what<00:08:06.960> it's<00:08:07.120> worth,<00:08:07.599> that's 00:08:07.830 --> 00:08:07.840 align:start position:0% hard. And for what it's worth, that's 00:08:07.840 --> 00:08:09.589 align:start position:0% hard. And for what it's worth, that's why<00:08:08.080> we<00:08:08.319> don't<00:08:08.479> sell<00:08:08.720> our<00:08:08.879> own<00:08:09.039> subscriptions 00:08:09.589 --> 00:08:09.599 align:start position:0% why we don't sell our own subscriptions 00:08:09.599 --> 00:08:12.390 align:start position:0% why we don't sell our own subscriptions in<00:08:09.840> T3<00:08:10.319> Code.<00:08:10.800> It<00:08:11.039> is<00:08:11.280> way<00:08:11.520> too<00:08:11.759> expensive<00:08:12.160> to 00:08:12.390 --> 00:08:12.400 align:start position:0% in T3 Code. It is way too expensive to 00:08:12.400 --> 00:08:14.550 align:start position:0% in T3 Code. It is way too expensive to hit<00:08:12.639> Cloud<00:08:12.960> models<00:08:13.280> over<00:08:13.599> API.<00:08:14.240> And<00:08:14.400> that's 00:08:14.550 --> 00:08:14.560 align:start position:0% hit Cloud models over API. And that's 00:08:14.560 --> 00:08:16.150 align:start position:0% hit Cloud models over API. And that's why<00:08:14.720> we<00:08:14.879> let<00:08:15.039> you<00:08:15.199> bring<00:08:15.360> your<00:08:15.520> own<00:08:15.680> sub.<00:08:16.000> So, 00:08:16.150 --> 00:08:16.160 align:start position:0% why we let you bring your own sub. So, 00:08:16.160 --> 00:08:17.830 align:start position:0% why we let you bring your own sub. So, if<00:08:16.319> you're<00:08:16.400> using<00:08:16.560> Codeex<00:08:17.039> with<00:08:17.120> the<00:08:17.280> OpenAI 00:08:17.830 --> 00:08:17.840 align:start position:0% if you're using Codeex with the OpenAI 00:08:17.840 --> 00:08:19.670 align:start position:0% if you're using Codeex with the OpenAI subscription<00:08:18.240> or<00:08:18.560> Cloud<00:08:18.879> with<00:08:18.960> an<00:08:19.199> anthropic 00:08:19.670 --> 00:08:19.680 align:start position:0% subscription or Cloud with an anthropic 00:08:19.680 --> 00:08:21.749 align:start position:0% subscription or Cloud with an anthropic subscription,<00:08:20.319> you<00:08:20.479> can<00:08:20.560> use<00:08:20.720> T3<00:08:21.199> Code<00:08:21.520> with 00:08:21.749 --> 00:08:21.759 align:start position:0% subscription, you can use T3 Code with 00:08:21.759 --> 00:08:23.270 align:start position:0% subscription, you can use T3 Code with either<00:08:22.080> of<00:08:22.240> those.<00:08:22.639> So,<00:08:22.720> if<00:08:22.879> you're<00:08:22.960> on<00:08:23.120> the 00:08:23.270 --> 00:08:23.280 align:start position:0% either of those. So, if you're on the 00:08:23.280 --> 00:08:25.029 align:start position:0% either of those. So, if you're on the $200<00:08:23.680> a<00:08:23.759> month<00:08:23.919> plan<00:08:24.080> for<00:08:24.160> Codex<00:08:24.479> and<00:08:24.560> the<00:08:24.720> $200 00:08:25.029 --> 00:08:25.039 align:start position:0% $200 a month plan for Codex and the $200 00:08:25.039 --> 00:08:26.469 align:start position:0% $200 a month plan for Codex and the $200 month<00:08:25.199> plan<00:08:25.440> for<00:08:25.520> Claude,<00:08:25.919> that's<00:08:26.160> like 00:08:26.469 --> 00:08:26.479 align:start position:0% month plan for Claude, that's like 00:08:26.479 --> 00:08:28.309 align:start position:0% month plan for Claude, that's like $10,000<00:08:27.120> of<00:08:27.280> inference<00:08:27.680> in<00:08:27.840> a<00:08:28.000> really<00:08:28.160> good 00:08:28.309 --> 00:08:28.319 align:start position:0% $10,000 of inference in a really good 00:08:28.319 --> 00:08:30.070 align:start position:0% $10,000 of inference in a really good interface.<00:08:28.879> Just<00:08:29.120> saying.<00:08:29.520> We<00:08:29.680> don't<00:08:29.840> charge 00:08:30.070 --> 00:08:30.080 align:start position:0% interface. Just saying. We don't charge 00:08:30.080 --> 00:08:31.830 align:start position:0% interface. Just saying. We don't charge money.<00:08:30.400> It's<00:08:30.639> free,<00:08:30.960> open<00:08:31.199> source.<00:08:31.599> You<00:08:31.680> can 00:08:31.830 --> 00:08:31.840 align:start position:0% money. It's free, open source. You can 00:08:31.840 --> 00:08:33.190 align:start position:0% money. It's free, open source. You can change<00:08:31.919> it<00:08:32.080> and<00:08:32.320> edit<00:08:32.560> and<00:08:32.719> do<00:08:32.800> whatever<00:08:33.039> you 00:08:33.190 --> 00:08:33.200 align:start position:0% change it and edit and do whatever you 00:08:33.200 --> 00:08:34.709 align:start position:0% change it and edit and do whatever you want<00:08:33.279> with<00:08:33.519> it.<00:08:33.760> But<00:08:33.919> if<00:08:34.080> you<00:08:34.159> have<00:08:34.240> these<00:08:34.479> subs 00:08:34.709 --> 00:08:34.719 align:start position:0% want with it. But if you have these subs 00:08:34.719 --> 00:08:36.230 align:start position:0% want with it. But if you have these subs and<00:08:34.880> you<00:08:34.959> want<00:08:35.039> a<00:08:35.200> UI<00:08:35.440> that<00:08:35.599> doesn't<00:08:35.839> suck, 00:08:36.230 --> 00:08:36.240 align:start position:0% and you want a UI that doesn't suck, 00:08:36.240 --> 00:08:38.389 align:start position:0% and you want a UI that doesn't suck, consider<00:08:36.640> T3<00:08:37.039> Code.<00:08:37.360> Just<00:08:37.519> just<00:08:37.760> saying.<00:08:38.240> But 00:08:38.389 --> 00:08:38.399 align:start position:0% consider T3 Code. Just just saying. But 00:08:38.399 --> 00:08:40.790 align:start position:0% consider T3 Code. Just just saying. But Cursor<00:08:38.959> can't<00:08:39.440> do<00:08:39.599> that.<00:08:40.080> And<00:08:40.159> there's<00:08:40.560> a 00:08:40.790 --> 00:08:40.800 align:start position:0% Cursor can't do that. And there's a 00:08:40.800 --> 00:08:43.029 align:start position:0% Cursor can't do that. And there's a bunch<00:08:41.120> of<00:08:41.360> reasons<00:08:41.680> why.<00:08:42.080> The<00:08:42.320> biggest<00:08:42.560> one<00:08:42.800> by 00:08:43.029 --> 00:08:43.039 align:start position:0% bunch of reasons why. The biggest one by 00:08:43.039 --> 00:08:44.550 align:start position:0% bunch of reasons why. The biggest one by far<00:08:43.200> is<00:08:43.360> that<00:08:43.519> they<00:08:43.680> have<00:08:43.760> to<00:08:43.919> maintain<00:08:44.159> a<00:08:44.399> good 00:08:44.550 --> 00:08:44.560 align:start position:0% far is that they have to maintain a good 00:08:44.560 --> 00:08:46.310 align:start position:0% far is that they have to maintain a good working<00:08:44.880> relationship<00:08:45.440> with<00:08:45.760> both<00:08:46.000> of<00:08:46.160> the 00:08:46.310 --> 00:08:46.320 align:start position:0% working relationship with both of the 00:08:46.320 --> 00:08:48.630 align:start position:0% working relationship with both of the labs<00:08:47.040> because<00:08:47.600> they<00:08:47.839> rely<00:08:48.080> on<00:08:48.240> them<00:08:48.399> so 00:08:48.630 --> 00:08:48.640 align:start position:0% labs because they rely on them so 00:08:48.640 --> 00:08:50.710 align:start position:0% labs because they rely on them so heavily<00:08:48.959> in<00:08:49.279> their<00:08:49.600> application<00:08:50.240> and<00:08:50.480> a<00:08:50.640> lot 00:08:50.710 --> 00:08:50.720 align:start position:0% heavily in their application and a lot 00:08:50.720 --> 00:08:52.070 align:start position:0% heavily in their application and a lot of<00:08:50.800> cursor's<00:08:51.279> revenue<00:08:51.680> comes<00:08:51.920> from 00:08:52.070 --> 00:08:52.080 align:start position:0% of cursor's revenue comes from 00:08:52.080 --> 00:08:53.350 align:start position:0% of cursor's revenue comes from subscriptions<00:08:52.560> that<00:08:52.720> they're<00:08:52.959> selling<00:08:53.120> to 00:08:53.350 --> 00:08:53.360 align:start position:0% subscriptions that they're selling to 00:08:53.360 --> 00:08:55.269 align:start position:0% subscriptions that they're selling to giant<00:08:53.680> enterprises<00:08:54.399> that<00:08:54.640> are<00:08:54.800> paying<00:08:55.040> for 00:08:55.269 --> 00:08:55.279 align:start position:0% giant enterprises that are paying for 00:08:55.279 --> 00:08:56.949 align:start position:0% giant enterprises that are paying for this<00:08:55.440> inference<00:08:56.000> through<00:08:56.240> that<00:08:56.480> cursor 00:08:56.949 --> 00:08:56.959 align:start position:0% this inference through that cursor 00:08:56.959 --> 00:08:58.790 align:start position:0% this inference through that cursor subscription.<00:08:57.760> So<00:08:58.000> they<00:08:58.240> need<00:08:58.399> ways<00:08:58.640> to 00:08:58.790 --> 00:08:58.800 align:start position:0% subscription. So they need ways to 00:08:58.800 --> 00:09:01.190 align:start position:0% subscription. So they need ways to reduce<00:08:59.279> costs<00:08:59.839> because<00:09:00.080> if<00:09:00.320> Enthropic<00:09:00.959> can 00:09:01.190 --> 00:09:01.200 align:start position:0% reduce costs because if Enthropic can 00:09:01.200 --> 00:09:04.470 align:start position:0% reduce costs because if Enthropic can charge<00:09:01.920> Cursor<00:09:02.640> $200<00:09:03.200> for<00:09:03.519> $200<00:09:03.839> of<00:09:04.000> inference 00:09:04.470 --> 00:09:04.480 align:start position:0% charge Cursor $200 for $200 of inference 00:09:04.480 --> 00:09:05.750 align:start position:0% charge Cursor $200 for $200 of inference and<00:09:04.640> then<00:09:04.800> they<00:09:04.959> have<00:09:05.040> to<00:09:05.200> try<00:09:05.279> and<00:09:05.440> find<00:09:05.600> a<00:09:05.680> way 00:09:05.750 --> 00:09:05.760 align:start position:0% and then they have to try and find a way 00:09:05.760 --> 00:09:07.750 align:start position:0% and then they have to try and find a way to<00:09:05.920> justify<00:09:06.240> that<00:09:06.399> to<00:09:06.560> their<00:09:06.720> users,<00:09:07.360> but<00:09:07.519> they 00:09:07.750 --> 00:09:07.760 align:start position:0% to justify that to their users, but they 00:09:07.760 --> 00:09:10.470 align:start position:0% to justify that to their users, but they go<00:09:07.839> and<00:09:08.080> resell<00:09:08.480> the<00:09:08.640> same<00:09:08.800> inference<00:09:09.279> for<00:09:09.680> 25x 00:09:10.470 --> 00:09:10.480 align:start position:0% go and resell the same inference for 25x 00:09:10.480 --> 00:09:11.829 align:start position:0% go and resell the same inference for 25x discount<00:09:10.959> to<00:09:11.200> people<00:09:11.360> on<00:09:11.600> their 00:09:11.829 --> 00:09:11.839 align:start position:0% discount to people on their 00:09:11.839 --> 00:09:13.269 align:start position:0% discount to people on their subscription.<00:09:12.560> You're<00:09:12.800> kind<00:09:12.959> of<00:09:13.040> just 00:09:13.269 --> 00:09:13.279 align:start position:0% subscription. You're kind of just 00:09:13.279 --> 00:09:14.870 align:start position:0% subscription. You're kind of just screwed.<00:09:13.839> And<00:09:14.000> I<00:09:14.160> get<00:09:14.240> it.<00:09:14.399> Like<00:09:14.640> I<00:09:14.800> don't<00:09:14.800> know 00:09:14.870 --> 00:09:14.880 align:start position:0% screwed. And I get it. Like I don't know 00:09:14.880 --> 00:09:16.310 align:start position:0% screwed. And I get it. Like I don't know how<00:09:15.040> we're<00:09:15.200> going<00:09:15.279> to<00:09:15.360> make<00:09:15.519> money<00:09:15.600> on<00:09:15.839> T3 00:09:16.310 --> 00:09:16.320 align:start position:0% how we're going to make money on T3 00:09:16.320 --> 00:09:18.150 align:start position:0% how we're going to make money on T3 code.<00:09:16.720> Like<00:09:16.959> we're<00:09:17.200> not<00:09:17.360> charging<00:09:17.760> subs.<00:09:18.000> It's 00:09:18.150 --> 00:09:18.160 align:start position:0% code. Like we're not charging subs. It's 00:09:18.160 --> 00:09:19.590 align:start position:0% code. Like we're not charging subs. It's kind<00:09:18.320> of<00:09:18.399> crazy<00:09:18.560> that<00:09:18.800> we<00:09:18.959> spent<00:09:19.120> all<00:09:19.279> of<00:09:19.360> this 00:09:19.590 --> 00:09:19.600 align:start position:0% kind of crazy that we spent all of this 00:09:19.600 --> 00:09:21.670 align:start position:0% kind of crazy that we spent all of this time<00:09:19.680> and<00:09:19.920> money<00:09:20.080> and<00:09:20.320> energy,<00:09:20.800> effort<00:09:21.200> and 00:09:21.670 --> 00:09:21.680 align:start position:0% time and money and energy, effort and 00:09:21.680 --> 00:09:24.070 align:start position:0% time and money and energy, effort and space<00:09:22.320> to<00:09:22.640> build<00:09:22.880> something<00:09:23.440> that<00:09:23.680> is<00:09:23.839> making 00:09:24.070 --> 00:09:24.080 align:start position:0% space to build something that is making 00:09:24.080 --> 00:09:26.310 align:start position:0% space to build something that is making it<00:09:24.240> so<00:09:24.399> OpenAI<00:09:25.040> and<00:09:25.279> Anthropic<00:09:25.920> sell<00:09:26.080> more 00:09:26.310 --> 00:09:26.320 align:start position:0% it so OpenAI and Anthropic sell more 00:09:26.320 --> 00:09:27.509 align:start position:0% it so OpenAI and Anthropic sell more subscriptions<00:09:26.720> and<00:09:26.880> we<00:09:26.959> don't<00:09:27.040> get<00:09:27.200> paid<00:09:27.360> a 00:09:27.509 --> 00:09:27.519 align:start position:0% subscriptions and we don't get paid a 00:09:27.519 --> 00:09:29.269 align:start position:0% subscriptions and we don't get paid a scent.<00:09:27.839> This<00:09:28.000> is<00:09:28.160> not<00:09:28.320> the<00:09:28.560> most<00:09:28.720> realistic 00:09:29.269 --> 00:09:29.279 align:start position:0% scent. This is not the most realistic 00:09:29.279 --> 00:09:31.430 align:start position:0% scent. This is not the most realistic path<00:09:29.680> as<00:09:29.920> a<00:09:30.080> business.<00:09:30.640> We<00:09:30.880> even<00:09:31.040> make<00:09:31.200> you 00:09:31.430 --> 00:09:31.440 align:start position:0% path as a business. We even make you 00:09:31.440 --> 00:09:33.990 align:start position:0% path as a business. We even make you install<00:09:31.839> the<00:09:32.080> official<00:09:32.480> cloud<00:09:32.720> and<00:09:32.959> Codex<00:09:33.440> CLI 00:09:33.990 --> 00:09:34.000 align:start position:0% install the official cloud and Codex CLI 00:09:34.000 --> 00:09:35.590 align:start position:0% install the official cloud and Codex CLI if<00:09:34.080> you<00:09:34.240> want<00:09:34.320> to<00:09:34.399> use<00:09:34.560> those.<00:09:34.800> Like<00:09:35.040> we<00:09:35.279> don't 00:09:35.590 --> 00:09:35.600 align:start position:0% if you want to use those. Like we don't 00:09:35.600 --> 00:09:37.829 align:start position:0% if you want to use those. Like we don't do<00:09:35.839> anything<00:09:36.160> over<00:09:36.480> API<00:09:37.120> or<00:09:37.279> with<00:09:37.519> our<00:09:37.680> own 00:09:37.829 --> 00:09:37.839 align:start position:0% do anything over API or with our own 00:09:37.839 --> 00:09:39.030 align:start position:0% do anything over API or with our own integration<00:09:38.240> or<00:09:38.399> anything<00:09:38.640> we<00:09:38.800> can<00:09:38.880> charge 00:09:39.030 --> 00:09:39.040 align:start position:0% integration or anything we can charge 00:09:39.040 --> 00:09:40.470 align:start position:0% integration or anything we can charge for<00:09:39.279> and<00:09:39.440> I<00:09:39.600> don't<00:09:39.680> even<00:09:39.839> know<00:09:39.920> how<00:09:40.160> we<00:09:40.320> would 00:09:40.470 --> 00:09:40.480 align:start position:0% for and I don't even know how we would 00:09:40.480 --> 00:09:42.070 align:start position:0% for and I don't even know how we would at<00:09:40.640> this<00:09:40.800> point.<00:09:41.040> We<00:09:41.279> have<00:09:41.440> things<00:09:41.680> we<00:09:41.839> want<00:09:41.920> to 00:09:42.070 --> 00:09:42.080 align:start position:0% at this point. We have things we want to 00:09:42.080 --> 00:09:43.269 align:start position:0% at this point. We have things we want to build<00:09:42.160> in<00:09:42.320> the<00:09:42.399> future<00:09:42.640> some<00:09:42.800> of<00:09:42.880> which<00:09:43.120> might 00:09:43.269 --> 00:09:43.279 align:start position:0% build in the future some of which might 00:09:43.279 --> 00:09:45.590 align:start position:0% build in the future some of which might have<00:09:43.440> to<00:09:43.600> cost<00:09:43.839> money<00:09:44.080> but<00:09:44.399> like<00:09:44.959> we<00:09:45.200> are<00:09:45.360> doing 00:09:45.590 --> 00:09:45.600 align:start position:0% have to cost money but like we are doing 00:09:45.600 --> 00:09:47.750 align:start position:0% have to cost money but like we are doing this<00:09:45.760> because<00:09:46.000> we<00:09:46.240> want<00:09:46.399> it.<00:09:46.959> cursor<00:09:47.440> needs<00:09:47.600> to 00:09:47.750 --> 00:09:47.760 align:start position:0% this because we want it. cursor needs to 00:09:47.760 --> 00:09:49.430 align:start position:0% this because we want it. cursor needs to make<00:09:48.000> billions<00:09:48.399> of<00:09:48.480> dollars<00:09:48.800> to<00:09:49.040> justify 00:09:49.430 --> 00:09:49.440 align:start position:0% make billions of dollars to justify 00:09:49.440 --> 00:09:51.590 align:start position:0% make billions of dollars to justify their<00:09:49.760> billions<00:09:50.160> of<00:09:50.320> dollar<00:09:50.640> valuation.<00:09:51.360> They 00:09:51.590 --> 00:09:51.600 align:start position:0% their billions of dollar valuation. They 00:09:51.600 --> 00:09:53.509 align:start position:0% their billions of dollar valuation. They need<00:09:51.760> to<00:09:51.920> be<00:09:52.000> able<00:09:52.160> to<00:09:52.320> make<00:09:52.480> a<00:09:52.720> profit<00:09:53.120> off<00:09:53.360> of 00:09:53.509 --> 00:09:53.519 align:start position:0% need to be able to make a profit off of 00:09:53.519 --> 00:09:55.590 align:start position:0% need to be able to make a profit off of the<00:09:53.680> inference<00:09:54.240> and<00:09:54.560> anthropic<00:09:55.040> is<00:09:55.200> making<00:09:55.360> it 00:09:55.590 --> 00:09:55.600 align:start position:0% the inference and anthropic is making it 00:09:55.600 --> 00:09:57.430 align:start position:0% the inference and anthropic is making it very<00:09:55.839> hard<00:09:56.000> for<00:09:56.160> them<00:09:56.320> to<00:09:56.480> do<00:09:56.640> that.<00:09:56.959> So<00:09:57.120> they 00:09:57.430 --> 00:09:57.440 align:start position:0% very hard for them to do that. So they 00:09:57.440 --> 00:09:59.509 align:start position:0% very hard for them to do that. So they need<00:09:57.600> a<00:09:57.760> wedge.<00:09:58.320> We<00:09:58.560> need<00:09:58.640> to<00:09:58.800> think<00:09:58.959> about 00:09:59.509 --> 00:09:59.519 align:start position:0% need a wedge. We need to think about 00:09:59.519 --> 00:10:01.350 align:start position:0% need a wedge. We need to think about leverage<00:10:00.000> for<00:10:00.240> a<00:10:00.480> second<00:10:00.640> here<00:10:01.040> because 00:10:01.350 --> 00:10:01.360 align:start position:0% leverage for a second here because 00:10:01.360 --> 00:10:03.269 align:start position:0% leverage for a second here because Anthropic<00:10:02.080> has<00:10:02.320> a<00:10:02.480> lot<00:10:02.560> of<00:10:02.720> leverage.<00:10:03.120> They 00:10:03.269 --> 00:10:03.279 align:start position:0% Anthropic has a lot of leverage. They 00:10:03.279 --> 00:10:05.030 align:start position:0% Anthropic has a lot of leverage. They have<00:10:03.360> the<00:10:03.519> model<00:10:03.839> everybody<00:10:04.240> wants<00:10:04.480> to<00:10:04.640> use. 00:10:05.030 --> 00:10:05.040 align:start position:0% have the model everybody wants to use. 00:10:05.040 --> 00:10:07.190 align:start position:0% have the model everybody wants to use. They<00:10:05.360> have<00:10:05.519> the<00:10:05.839> APIs<00:10:06.399> everybody<00:10:06.720> has<00:10:06.880> to<00:10:07.040> use 00:10:07.190 --> 00:10:07.200 align:start position:0% They have the APIs everybody has to use 00:10:07.200 --> 00:10:09.190 align:start position:0% They have the APIs everybody has to use to<00:10:07.440> use<00:10:07.600> them<00:10:07.839> and<00:10:08.080> as<00:10:08.320> such<00:10:08.640> everybody<00:10:08.880> has<00:10:09.120> to 00:10:09.190 --> 00:10:09.200 align:start position:0% to use them and as such everybody has to 00:10:09.200 --> 00:10:11.190 align:start position:0% to use them and as such everybody has to pay<00:10:09.440> them<00:10:09.760> to<00:10:10.160> get<00:10:10.320> the<00:10:10.480> thing<00:10:10.640> they<00:10:10.880> want 00:10:11.190 --> 00:10:11.200 align:start position:0% pay them to get the thing they want 00:10:11.200 --> 00:10:13.350 align:start position:0% pay them to get the thing they want which<00:10:11.360> is<00:10:11.519> inference.<00:10:12.320> So<00:10:12.480> what<00:10:12.720> does<00:10:12.880> cursor 00:10:13.350 --> 00:10:13.360 align:start position:0% which is inference. So what does cursor 00:10:13.360 --> 00:10:15.990 align:start position:0% which is inference. So what does cursor have?<00:10:13.680> Cursor<00:10:14.160> has<00:10:14.560> a<00:10:14.800> large<00:10:15.040> install<00:10:15.360> base 00:10:15.990 --> 00:10:16.000 align:start position:0% have? Cursor has a large install base 00:10:16.000 --> 00:10:17.670 align:start position:0% have? Cursor has a large install base which<00:10:16.240> helps<00:10:16.399> a<00:10:16.640> ton.<00:10:16.880> It<00:10:17.040> means<00:10:17.279> they<00:10:17.519> can 00:10:17.670 --> 00:10:17.680 align:start position:0% which helps a ton. It means they can 00:10:17.680 --> 00:10:19.269 align:start position:0% which helps a ton. It means they can charge<00:10:18.000> these<00:10:18.240> people<00:10:18.399> money<00:10:18.640> and<00:10:18.880> make<00:10:19.040> some 00:10:19.269 --> 00:10:19.279 align:start position:0% charge these people money and make some 00:10:19.279 --> 00:10:20.710 align:start position:0% charge these people money and make some money.<00:10:19.519> But<00:10:19.680> cursor<00:10:20.079> has<00:10:20.320> something<00:10:20.480> else 00:10:20.710 --> 00:10:20.720 align:start position:0% money. But cursor has something else 00:10:20.720 --> 00:10:22.790 align:start position:0% money. But cursor has something else that's<00:10:21.040> really<00:10:21.279> valuable.<00:10:21.760> Chat<00:10:22.160> histories. 00:10:22.790 --> 00:10:22.800 align:start position:0% that's really valuable. Chat histories. 00:10:22.800 --> 00:10:24.630 align:start position:0% that's really valuable. Chat histories. When<00:10:22.959> you<00:10:23.120> do<00:10:23.279> an<00:10:23.440> API<00:10:23.920> request<00:10:24.320> through 00:10:24.630 --> 00:10:24.640 align:start position:0% When you do an API request through 00:10:24.640 --> 00:10:26.710 align:start position:0% When you do an API request through something<00:10:24.959> like<00:10:25.360> cursor,<00:10:26.000> when<00:10:26.160> you<00:10:26.320> make<00:10:26.480> a 00:10:26.710 --> 00:10:26.720 align:start position:0% something like cursor, when you make a 00:10:26.720 --> 00:10:28.550 align:start position:0% something like cursor, when you make a prompt<00:10:26.959> in<00:10:27.120> a<00:10:27.279> thing<00:10:27.440> like<00:10:27.600> cursor<00:10:28.079> that<00:10:28.320> is<00:10:28.480> to 00:10:28.550 --> 00:10:28.560 align:start position:0% prompt in a thing like cursor that is to 00:10:28.560 --> 00:10:30.230 align:start position:0% prompt in a thing like cursor that is to go<00:10:28.720> through<00:10:28.880> their<00:10:29.120> servers<00:10:29.600> and<00:10:29.760> that<00:10:30.000> chat 00:10:30.230 --> 00:10:30.240 align:start position:0% go through their servers and that chat 00:10:30.240 --> 00:10:32.710 align:start position:0% go through their servers and that chat history<00:10:30.959> is<00:10:31.360> a<00:10:31.600> thing<00:10:31.760> that<00:10:32.000> exists.<00:10:32.480> If<00:10:32.640> you 00:10:32.710 --> 00:10:32.720 align:start position:0% history is a thing that exists. If you 00:10:32.720 --> 00:10:34.150 align:start position:0% history is a thing that exists. If you don't<00:10:32.800> click<00:10:32.959> the<00:10:33.120> switchite<00:10:33.600> of<00:10:33.680> cursor<00:10:34.000> that 00:10:34.150 --> 00:10:34.160 align:start position:0% don't click the switchite of cursor that 00:10:34.160 --> 00:10:36.150 align:start position:0% don't click the switchite of cursor that says<00:10:34.399> don't<00:10:34.640> train<00:10:34.800> on<00:10:34.959> my<00:10:35.120> data,<00:10:35.519> that<00:10:35.839> data 00:10:36.150 --> 00:10:36.160 align:start position:0% says don't train on my data, that data 00:10:36.160 --> 00:10:38.150 align:start position:0% says don't train on my data, that data all<00:10:36.480> goes<00:10:36.800> to<00:10:37.120> cursor.<00:10:37.680> So<00:10:37.839> I<00:10:38.000> would 00:10:38.150 --> 00:10:38.160 align:start position:0% all goes to cursor. So I would 00:10:38.160 --> 00:10:41.269 align:start position:0% all goes to cursor. So I would personally<00:10:38.640> estimate<00:10:39.360> that<00:10:40.000> maybe<00:10:40.800> half<00:10:41.040> or 00:10:41.269 --> 00:10:41.279 align:start position:0% personally estimate that maybe half or 00:10:41.279 --> 00:10:44.310 align:start position:0% personally estimate that maybe half or so<00:10:41.760> of<00:10:42.079> all<00:10:42.320> the<00:10:42.560> anthropic<00:10:43.120> prompts<00:10:43.760> that<00:10:44.079> go 00:10:44.310 --> 00:10:44.320 align:start position:0% so of all the anthropic prompts that go 00:10:44.320 --> 00:10:46.310 align:start position:0% so of all the anthropic prompts that go through<00:10:44.560> cursor<00:10:45.120> are<00:10:45.440> going<00:10:45.600> through<00:10:45.839> it<00:10:45.920> in<00:10:46.160> a 00:10:46.310 --> 00:10:46.320 align:start position:0% through cursor are going through it in a 00:10:46.320 --> 00:10:48.470 align:start position:0% through cursor are going through it in a way<00:10:46.399> where<00:10:46.640> they're<00:10:46.959> earmarked<00:10:47.600> as<00:10:48.000> usable 00:10:48.470 --> 00:10:48.480 align:start position:0% way where they're earmarked as usable 00:10:48.480 --> 00:10:49.829 align:start position:0% way where they're earmarked as usable for<00:10:48.720> training.<00:10:49.200> That<00:10:49.360> means<00:10:49.519> they<00:10:49.680> have<00:10:49.760> a 00:10:49.829 --> 00:10:49.839 align:start position:0% for training. That means they have a 00:10:49.839 --> 00:10:52.389 align:start position:0% for training. That means they have a shitload<00:10:50.320> of<00:10:50.720> history,<00:10:51.200> a<00:10:51.360> shitload<00:10:51.680> of<00:10:51.920> data 00:10:52.389 --> 00:10:52.399 align:start position:0% shitload of history, a shitload of data 00:10:52.399 --> 00:10:54.870 align:start position:0% shitload of history, a shitload of data of<00:10:52.720> existing<00:10:53.519> models<00:10:53.839> that<00:10:54.079> are<00:10:54.240> really<00:10:54.480> good 00:10:54.870 --> 00:10:54.880 align:start position:0% of existing models that are really good 00:10:54.880 --> 00:10:57.350 align:start position:0% of existing models that are really good working<00:10:55.200> in<00:10:55.360> these<00:10:55.680> realworld<00:10:56.399> code<00:10:56.720> bases 00:10:57.350 --> 00:10:57.360 align:start position:0% working in these realworld code bases 00:10:57.360 --> 00:11:00.230 align:start position:0% working in these realworld code bases that<00:10:57.600> a<00:10:57.920> lot<00:10:58.079> of<00:10:58.320> people<00:10:58.959> just<00:10:59.360> do<00:10:59.600> not<00:10:59.839> have. 00:11:00.230 --> 00:11:00.240 align:start position:0% that a lot of people just do not have. 00:11:00.240 --> 00:11:02.550 align:start position:0% that a lot of people just do not have. They<00:11:00.399> have<00:11:00.640> the<00:11:00.880> craziest<00:11:01.440> wide<00:11:01.760> variety<00:11:02.320> of 00:11:02.550 --> 00:11:02.560 align:start position:0% They have the craziest wide variety of 00:11:02.560 --> 00:11:04.870 align:start position:0% They have the craziest wide variety of data<00:11:02.959> and<00:11:03.360> chat<00:11:03.680> histories<00:11:04.079> they<00:11:04.320> can<00:11:04.480> use<00:11:04.640> to 00:11:04.870 --> 00:11:04.880 align:start position:0% data and chat histories they can use to 00:11:04.880 --> 00:11:06.389 align:start position:0% data and chat histories they can use to do<00:11:05.040> reinforcement<00:11:05.680> learning<00:11:05.920> and<00:11:06.079> training 00:11:06.389 --> 00:11:06.399 align:start position:0% do reinforcement learning and training 00:11:06.399 --> 00:11:08.150 align:start position:0% do reinforcement learning and training on<00:11:06.640> the<00:11:06.800> models.<00:11:07.360> You<00:11:07.519> guys<00:11:07.680> might<00:11:07.920> remember 00:11:08.150 --> 00:11:08.160 align:start position:0% on the models. You guys might remember 00:11:08.160 --> 00:11:10.310 align:start position:0% on the models. You guys might remember this<00:11:08.480> article<00:11:08.800> from<00:11:09.120> earlier<00:11:09.519> this<00:11:09.839> year<00:11:10.000> from 00:11:10.310 --> 00:11:10.320 align:start position:0% this article from earlier this year from 00:11:10.320 --> 00:11:11.990 align:start position:0% this article from earlier this year from Anthropic<00:11:10.959> where<00:11:11.200> they<00:11:11.360> talked<00:11:11.600> about 00:11:11.990 --> 00:11:12.000 align:start position:0% Anthropic where they talked about 00:11:12.000 --> 00:11:13.750 align:start position:0% Anthropic where they talked about detecting<00:11:12.480> and<00:11:12.720> preventing<00:11:13.120> distillation 00:11:13.750 --> 00:11:13.760 align:start position:0% detecting and preventing distillation 00:11:13.760 --> 00:11:15.990 align:start position:0% detecting and preventing distillation attacks<00:11:14.320> where<00:11:14.560> they<00:11:14.800> claimed<00:11:15.200> that<00:11:15.440> labs 00:11:15.990 --> 00:11:16.000 align:start position:0% attacks where they claimed that labs 00:11:16.000 --> 00:11:20.389 align:start position:0% attacks where they claimed that labs like<00:11:16.240> Deepseek<00:11:17.200> stole<00:11:18.160> 150,000<00:11:19.279> exchanges<00:11:20.240> as 00:11:20.389 --> 00:11:20.399 align:start position:0% like Deepseek stole 150,000 exchanges as 00:11:20.399 --> 00:11:22.230 align:start position:0% like Deepseek stole 150,000 exchanges as in<00:11:20.560> back<00:11:20.800> and<00:11:20.959> forths<00:11:21.279> with<00:11:21.440> the<00:11:21.600> model<00:11:22.000> in 00:11:22.230 --> 00:11:22.240 align:start position:0% in back and forths with the model in 00:11:22.240 --> 00:11:24.069 align:start position:0% in back and forths with the model in order<00:11:22.399> to<00:11:22.640> train<00:11:22.880> and<00:11:23.120> refine<00:11:23.519> their<00:11:23.839> own 00:11:24.069 --> 00:11:24.079 align:start position:0% order to train and refine their own 00:11:24.079 --> 00:11:26.150 align:start position:0% order to train and refine their own model.<00:11:24.640> I<00:11:24.880> can't<00:11:25.120> properly<00:11:25.519> put<00:11:25.680> into<00:11:25.920> words 00:11:26.150 --> 00:11:26.160 align:start position:0% model. I can't properly put into words 00:11:26.160 --> 00:11:28.230 align:start position:0% model. I can't properly put into words how<00:11:26.399> absurd<00:11:26.880> this<00:11:27.120> is,<00:11:27.440> particularly<00:11:28.000> this 00:11:28.230 --> 00:11:28.240 align:start position:0% how absurd this is, particularly this 00:11:28.240 --> 00:11:30.949 align:start position:0% how absurd this is, particularly this number.<00:11:28.800> 150,000<00:11:29.360> exchanges<00:11:30.079> is<00:11:30.480> nothing. 00:11:30.949 --> 00:11:30.959 align:start position:0% number. 150,000 exchanges is nothing. 00:11:30.959 --> 00:11:32.710 align:start position:0% number. 150,000 exchanges is nothing. Since<00:11:31.200> I<00:11:31.360> started<00:11:31.600> recording<00:11:31.920> this<00:11:32.160> video, 00:11:32.710 --> 00:11:32.720 align:start position:0% Since I started recording this video, 00:11:32.720 --> 00:11:35.430 align:start position:0% Since I started recording this video, cursor's<00:11:33.279> probably<00:11:33.519> done<00:11:34.000> 150,000<00:11:34.480> exchanges 00:11:35.430 --> 00:11:35.440 align:start position:0% cursor's probably done 150,000 exchanges 00:11:35.440 --> 00:11:38.150 align:start position:0% cursor's probably done 150,000 exchanges 150,000<00:11:36.079> times.<00:11:36.720> This<00:11:36.880> is<00:11:37.040> not<00:11:37.200> much<00:11:37.519> data<00:11:37.920> at 00:11:38.150 --> 00:11:38.160 align:start position:0% 150,000 times. This is not much data at 00:11:38.160 --> 00:11:40.150 align:start position:0% 150,000 times. This is not much data at all.<00:11:38.480> T3<00:11:38.959> Chat<00:11:39.279> does<00:11:39.519> more<00:11:39.600> than<00:11:39.760> this<00:11:39.839> with 00:11:40.150 --> 00:11:40.160 align:start position:0% all. T3 Chat does more than this with 00:11:40.160 --> 00:11:42.710 align:start position:0% all. T3 Chat does more than this with anthropic<00:11:40.560> models<00:11:41.279> daily.<00:11:42.079> This<00:11:42.240> is<00:11:42.399> not<00:11:42.560> a 00:11:42.710 --> 00:11:42.720 align:start position:0% anthropic models daily. This is not a 00:11:42.720 --> 00:11:44.470 align:start position:0% anthropic models daily. This is not a number<00:11:42.959> that<00:11:43.200> matters.<00:11:43.839> I<00:11:44.079> do<00:11:44.240> want<00:11:44.320> to<00:11:44.399> be 00:11:44.470 --> 00:11:44.480 align:start position:0% number that matters. I do want to be 00:11:44.480 --> 00:11:46.150 align:start position:0% number that matters. I do want to be very<00:11:44.720> very<00:11:44.959> clear<00:11:45.120> cuz<00:11:45.279> I<00:11:45.440> see<00:11:45.600> people<00:11:45.839> already 00:11:46.150 --> 00:11:46.160 align:start position:0% very very clear cuz I see people already 00:11:46.160 --> 00:11:48.389 align:start position:0% very very clear cuz I see people already confused.<00:11:46.800> Cursor<00:11:47.279> is<00:11:47.519> not<00:11:47.680> using<00:11:47.920> your<00:11:48.160> data 00:11:48.389 --> 00:11:48.399 align:start position:0% confused. Cursor is not using your data 00:11:48.399 --> 00:11:50.310 align:start position:0% confused. Cursor is not using your data if<00:11:48.640> you<00:11:48.720> have<00:11:48.880> privacy<00:11:49.279> mode<00:11:49.519> on.<00:11:49.760> I<00:11:50.000> know<00:11:50.160> this 00:11:50.310 --> 00:11:50.320 align:start position:0% if you have privacy mode on. I know this 00:11:50.320 --> 00:11:52.069 align:start position:0% if you have privacy mode on. I know this for<00:11:50.480> a<00:11:50.720> fact<00:11:50.959> because<00:11:51.279> I<00:11:51.440> had<00:11:51.600> issues<00:11:51.920> with 00:11:52.069 --> 00:11:52.079 align:start position:0% for a fact because I had issues with 00:11:52.079 --> 00:11:53.990 align:start position:0% for a fact because I had issues with cursor<00:11:52.399> when<00:11:52.560> I<00:11:52.720> had<00:11:52.880> privacy<00:11:53.279> mode<00:11:53.440> on<00:11:53.760> and 00:11:53.990 --> 00:11:54.000 align:start position:0% cursor when I had privacy mode on and 00:11:54.000 --> 00:11:56.230 align:start position:0% cursor when I had privacy mode on and they<00:11:54.160> literally<00:11:54.720> could<00:11:55.040> not<00:11:55.279> debug<00:11:55.680> with<00:11:55.920> me 00:11:56.230 --> 00:11:56.240 align:start position:0% they literally could not debug with me 00:11:56.240 --> 00:11:58.230 align:start position:0% they literally could not debug with me because<00:11:56.480> they<00:11:56.720> had<00:11:56.880> no<00:11:57.040> way<00:11:57.279> to<00:11:57.440> access 00:11:58.230 --> 00:11:58.240 align:start position:0% because they had no way to access 00:11:58.240 --> 00:12:00.310 align:start position:0% because they had no way to access anything.<00:11:58.800> I<00:11:59.040> ended<00:11:59.200> up<00:11:59.360> turning<00:11:59.600> off<00:11:59.839> privacy 00:12:00.310 --> 00:12:00.320 align:start position:0% anything. I ended up turning off privacy 00:12:00.320 --> 00:12:02.150 align:start position:0% anything. I ended up turning off privacy mode<00:12:00.640> just<00:12:00.880> so<00:12:01.040> I<00:12:01.200> could<00:12:01.360> work<00:12:01.519> with<00:12:01.680> the<00:12:01.920> team 00:12:02.150 --> 00:12:02.160 align:start position:0% mode just so I could work with the team 00:12:02.160 --> 00:12:03.990 align:start position:0% mode just so I could work with the team more.<00:12:02.640> So,<00:12:02.720> our<00:12:02.880> friends<00:12:03.040> at<00:12:03.200> Anthropic<00:12:03.760> seem 00:12:03.990 --> 00:12:04.000 align:start position:0% more. So, our friends at Anthropic seem 00:12:04.000 --> 00:12:06.389 align:start position:0% more. So, our friends at Anthropic seem to<00:12:04.079> think<00:12:04.560> 150,000<00:12:05.200> exchanges<00:12:06.000> with<00:12:06.240> an 00:12:06.389 --> 00:12:06.399 align:start position:0% to think 150,000 exchanges with an 00:12:06.399 --> 00:12:08.949 align:start position:0% to think 150,000 exchanges with an anthropic<00:12:06.959> model<00:12:07.200> is<00:12:07.440> enough<00:12:07.600> for<00:12:07.839> DeepS<00:12:08.639> to 00:12:08.949 --> 00:12:08.959 align:start position:0% anthropic model is enough for DeepS to 00:12:08.959 --> 00:12:10.550 align:start position:0% anthropic model is enough for DeepS to steal<00:12:09.279> their<00:12:09.519> reasoning<00:12:10.000> capabilities 00:12:10.550 --> 00:12:10.560 align:start position:0% steal their reasoning capabilities 00:12:10.560 --> 00:12:12.550 align:start position:0% steal their reasoning capabilities across<00:12:10.880> diverse<00:12:11.279> tasks.<00:12:12.000> How<00:12:12.160> much<00:12:12.320> do<00:12:12.399> you 00:12:12.550 --> 00:12:12.560 align:start position:0% across diverse tasks. How much do you 00:12:12.560 --> 00:12:14.629 align:start position:0% across diverse tasks. How much do you think<00:12:12.639> you<00:12:12.800> can<00:12:12.959> do<00:12:13.519> with<00:12:14.079> hundreds<00:12:14.480> of 00:12:14.629 --> 00:12:14.639 align:start position:0% think you can do with hundreds of 00:12:14.639 --> 00:12:17.350 align:start position:0% think you can do with hundreds of thousands<00:12:15.120> of<00:12:15.360> times<00:12:15.920> more<00:12:16.320> data?<00:12:17.040> Because 00:12:17.350 --> 00:12:17.360 align:start position:0% thousands of times more data? Because 00:12:17.360 --> 00:12:19.990 align:start position:0% thousands of times more data? Because obviously<00:12:17.760> cursor<00:12:18.240> has<00:12:18.720> insane<00:12:19.200> amounts<00:12:19.839> as 00:12:19.990 --> 00:12:20.000 align:start position:0% obviously cursor has insane amounts as 00:12:20.000 --> 00:12:21.750 align:start position:0% obviously cursor has insane amounts as well<00:12:20.160> as<00:12:20.399> all<00:12:20.560> of<00:12:20.720> their<00:12:20.880> own<00:12:21.040> internal<00:12:21.440> work. 00:12:21.750 --> 00:12:21.760 align:start position:0% well as all of their own internal work. 00:12:21.760 --> 00:12:23.750 align:start position:0% well as all of their own internal work. I<00:12:21.920> would<00:12:22.000> bet<00:12:22.160> that<00:12:22.399> employees<00:12:22.800> at<00:12:22.959> Cursor 00:12:23.750 --> 00:12:23.760 align:start position:0% I would bet that employees at Cursor 00:12:23.760 --> 00:12:26.870 align:start position:0% I would bet that employees at Cursor using<00:12:24.079> Cursor<00:12:24.560> with<00:12:24.720> Opus<00:12:25.279> alone<00:12:26.000> is<00:12:26.320> way<00:12:26.639> over 00:12:26.870 --> 00:12:26.880 align:start position:0% using Cursor with Opus alone is way over 00:12:26.880 --> 00:12:28.710 align:start position:0% using Cursor with Opus alone is way over this<00:12:27.120> number.<00:12:27.600> They<00:12:27.839> have<00:12:28.000> a<00:12:28.160> lot<00:12:28.240> of<00:12:28.399> data. 00:12:28.710 --> 00:12:28.720 align:start position:0% this number. They have a lot of data. 00:12:28.720 --> 00:12:30.069 align:start position:0% this number. They have a lot of data. And<00:12:28.880> while<00:12:29.040> reinforcement<00:12:29.600> learning<00:12:29.839> is<00:12:30.000> far 00:12:30.069 --> 00:12:30.079 align:start position:0% And while reinforcement learning is far 00:12:30.079 --> 00:12:31.829 align:start position:0% And while reinforcement learning is far from<00:12:30.240> a<00:12:30.399> perfect<00:12:30.639> technique,<00:12:31.120> it<00:12:31.360> has<00:12:31.519> proven 00:12:31.829 --> 00:12:31.839 align:start position:0% from a perfect technique, it has proven 00:12:31.839 --> 00:12:33.990 align:start position:0% from a perfect technique, it has proven to<00:12:32.079> work<00:12:32.320> really<00:12:32.720> well<00:12:33.040> in<00:12:33.279> these<00:12:33.600> types<00:12:33.839> of 00:12:33.990 --> 00:12:34.000 align:start position:0% to work really well in these types of 00:12:34.000 --> 00:12:35.750 align:start position:0% to work really well in these types of scenarios.<00:12:34.560> When<00:12:34.720> you<00:12:34.880> have<00:12:34.959> a<00:12:35.120> ton<00:12:35.279> of<00:12:35.519> good 00:12:35.750 --> 00:12:35.760 align:start position:0% scenarios. When you have a ton of good 00:12:35.760 --> 00:12:38.230 align:start position:0% scenarios. When you have a ton of good or<00:12:36.079> good<00:12:36.240> enough<00:12:36.639> data<00:12:37.120> that<00:12:37.440> shows<00:12:37.680> what<00:12:37.920> a 00:12:38.230 --> 00:12:38.240 align:start position:0% or good enough data that shows what a 00:12:38.240 --> 00:12:40.550 align:start position:0% or good enough data that shows what a successful<00:12:38.880> run<00:12:39.200> looks<00:12:39.519> like,<00:12:40.160> especially<00:12:40.399> if 00:12:40.550 --> 00:12:40.560 align:start position:0% successful run looks like, especially if 00:12:40.560 --> 00:12:42.150 align:start position:0% successful run looks like, especially if you<00:12:40.720> can<00:12:40.800> verify<00:12:41.120> the<00:12:41.360> results<00:12:41.519> at<00:12:41.760> the<00:12:41.839> end<00:12:42.000> in 00:12:42.150 --> 00:12:42.160 align:start position:0% you can verify the results at the end in 00:12:42.160 --> 00:12:44.470 align:start position:0% you can verify the results at the end in some<00:12:42.399> way,<00:12:42.720> you<00:12:42.959> can<00:12:43.040> now<00:12:43.440> train<00:12:43.760> models<00:12:44.160> with 00:12:44.470 --> 00:12:44.480 align:start position:0% some way, you can now train models with 00:12:44.480 --> 00:12:46.870 align:start position:0% some way, you can now train models with that<00:12:45.040> after<00:12:45.440> they're<00:12:45.760> already<00:12:46.079> trained.<00:12:46.639> The 00:12:46.870 --> 00:12:46.880 align:start position:0% that after they're already trained. The 00:12:46.880 --> 00:12:48.389 align:start position:0% that after they're already trained. The pre-training<00:12:47.440> kind<00:12:47.600> of<00:12:47.760> gives<00:12:47.920> the<00:12:48.079> models 00:12:48.389 --> 00:12:48.399 align:start position:0% pre-training kind of gives the models 00:12:48.399 --> 00:12:50.550 align:start position:0% pre-training kind of gives the models their<00:12:48.639> intelligence<00:12:49.360> and<00:12:49.760> knowledge.<00:12:50.399> And 00:12:50.550 --> 00:12:50.560 align:start position:0% their intelligence and knowledge. And 00:12:50.560 --> 00:12:52.230 align:start position:0% their intelligence and knowledge. And this<00:12:50.800> type<00:12:50.959> of<00:12:51.120> reinforcement<00:12:51.760> learning<00:12:52.000> and 00:12:52.230 --> 00:12:52.240 align:start position:0% this type of reinforcement learning and 00:12:52.240 --> 00:12:53.990 align:start position:0% this type of reinforcement learning and post-training<00:12:52.959> gives<00:12:53.279> them<00:12:53.600> better 00:12:53.990 --> 00:12:54.000 align:start position:0% post-training gives them better 00:12:54.000 --> 00:12:56.949 align:start position:0% post-training gives them better behaviors<00:12:54.800> and<00:12:55.440> just<00:12:55.760> overall<00:12:56.399> more<00:12:56.720> likely 00:12:56.949 --> 00:12:56.959 align:start position:0% behaviors and just overall more likely 00:12:56.959 --> 00:12:58.470 align:start position:0% behaviors and just overall more likely to<00:12:57.200> comply<00:12:57.519> with<00:12:57.680> the<00:12:57.839> systems<00:12:58.079> that<00:12:58.320> they 00:12:58.470 --> 00:12:58.480 align:start position:0% to comply with the systems that they 00:12:58.480 --> 00:13:00.069 align:start position:0% to comply with the systems that they work<00:12:58.639> in.<00:12:59.040> And<00:12:59.200> that's<00:12:59.360> where<00:12:59.519> we<00:12:59.680> get<00:12:59.839> to 00:13:00.069 --> 00:13:00.079 align:start position:0% work in. And that's where we get to 00:13:00.079 --> 00:13:02.470 align:start position:0% work in. And that's where we get to today's<00:13:00.560> drama,<00:13:01.120> which<00:13:01.360> is<00:13:01.680> the<00:13:01.920> involvement 00:13:02.470 --> 00:13:02.480 align:start position:0% today's drama, which is the involvement 00:13:02.480 --> 00:13:04.949 align:start position:0% today's drama, which is the involvement with<00:13:02.959> Kimmy.<00:13:03.519> As<00:13:03.680> I<00:13:03.839> mentioned<00:13:04.160> in<00:13:04.399> the<00:13:04.560> intro, 00:13:04.949 --> 00:13:04.959 align:start position:0% with Kimmy. As I mentioned in the intro, 00:13:04.959 --> 00:13:07.269 align:start position:0% with Kimmy. As I mentioned in the intro, this<00:13:05.200> started<00:13:05.600> because<00:13:06.079> Finn<00:13:06.639> found<00:13:06.880> in<00:13:07.120> the 00:13:07.269 --> 00:13:07.279 align:start position:0% this started because Finn found in the 00:13:07.279 --> 00:13:10.870 align:start position:0% this started because Finn found in the APIs<00:13:08.160> that<00:13:08.399> the<00:13:08.560> model<00:13:08.800> was<00:13:09.040> named<00:13:09.440> Kimmy<00:13:09.920> K2.5 00:13:10.870 --> 00:13:10.880 align:start position:0% APIs that the model was named Kimmy K2.5 00:13:10.880 --> 00:13:14.550 align:start position:0% APIs that the model was named Kimmy K2.5 reinforcement<00:13:11.519> learning<00:13:12.000> 0317<00:13:12.959> S515<00:13:14.000> fast. 00:13:14.550 --> 00:13:14.560 align:start position:0% reinforcement learning 0317 S515 fast. 00:13:14.560 --> 00:13:16.069 align:start position:0% reinforcement learning 0317 S515 fast. So<00:13:14.720> what<00:13:14.959> happened<00:13:15.120> here?<00:13:15.519> Well,<00:13:15.760> the<00:13:15.920> first 00:13:16.069 --> 00:13:16.079 align:start position:0% So what happened here? Well, the first 00:13:16.079 --> 00:13:18.069 align:start position:0% So what happened here? Well, the first thing<00:13:16.160> that<00:13:16.399> happened<00:13:16.720> was<00:13:17.040> an<00:13:17.360> employee<00:13:17.760> at 00:13:18.069 --> 00:13:18.079 align:start position:0% thing that happened was an employee at 00:13:18.079 --> 00:13:20.069 align:start position:0% thing that happened was an employee at Kimmy<00:13:18.480> and<00:13:18.720> Moonshot<00:13:19.279> being<00:13:19.519> very<00:13:19.760> confused 00:13:20.069 --> 00:13:20.079 align:start position:0% Kimmy and Moonshot being very confused 00:13:20.079 --> 00:13:20.949 align:start position:0% Kimmy and Moonshot being very confused because<00:13:20.240> they<00:13:20.320> had<00:13:20.480> not<00:13:20.560> heard<00:13:20.720> anything 00:13:20.949 --> 00:13:20.959 align:start position:0% because they had not heard anything 00:13:20.959 --> 00:13:22.710 align:start position:0% because they had not heard anything about<00:13:21.120> this.<00:13:21.600> This<00:13:21.760> is<00:13:21.839> an<00:13:22.000> important<00:13:22.320> detail 00:13:22.710 --> 00:13:22.720 align:start position:0% about this. This is an important detail 00:13:22.720 --> 00:13:24.550 align:start position:0% about this. This is an important detail because<00:13:22.880> it<00:13:23.120> means<00:13:23.440> that<00:13:23.680> the<00:13:23.920> knowledge<00:13:24.320> that 00:13:24.550 --> 00:13:24.560 align:start position:0% because it means that the knowledge that 00:13:24.560 --> 00:13:26.629 align:start position:0% because it means that the knowledge that this<00:13:24.800> model<00:13:25.040> was<00:13:25.279> being<00:13:25.440> used<00:13:25.680> by<00:13:25.839> cursor<00:13:26.399> was 00:13:26.629 --> 00:13:26.639 align:start position:0% this model was being used by cursor was 00:13:26.639 --> 00:13:29.269 align:start position:0% this model was being used by cursor was not<00:13:26.959> very<00:13:27.360> prolific<00:13:28.079> within<00:13:28.480> the<00:13:28.720> Moonshot 00:13:29.269 --> 00:13:29.279 align:start position:0% not very prolific within the Moonshot 00:13:29.279 --> 00:13:30.949 align:start position:0% not very prolific within the Moonshot team.<00:13:29.680> This<00:13:29.920> is<00:13:30.079> particularly<00:13:30.720> important 00:13:30.949 --> 00:13:30.959 align:start position:0% team. This is particularly important 00:13:30.959 --> 00:13:32.870 align:start position:0% team. This is particularly important because<00:13:31.200> of<00:13:31.360> the<00:13:31.519> modified<00:13:32.079> MIT<00:13:32.560> license 00:13:32.870 --> 00:13:32.880 align:start position:0% because of the modified MIT license 00:13:32.880 --> 00:13:35.190 align:start position:0% because of the modified MIT license they're<00:13:33.200> using.<00:13:33.839> Our<00:13:34.160> only<00:13:34.480> modification 00:13:35.190 --> 00:13:35.200 align:start position:0% they're using. Our only modification 00:13:35.200 --> 00:13:37.269 align:start position:0% they're using. Our only modification part<00:13:35.440> is<00:13:35.680> that<00:13:35.920> if<00:13:36.160> the<00:13:36.399> software<00:13:36.800> or<00:13:37.120> any 00:13:37.269 --> 00:13:37.279 align:start position:0% part is that if the software or any 00:13:37.279 --> 00:13:39.829 align:start position:0% part is that if the software or any derivative<00:13:37.760> works<00:13:38.079> thereof<00:13:38.639> is<00:13:38.880> used<00:13:39.279> for<00:13:39.600> any 00:13:39.829 --> 00:13:39.839 align:start position:0% derivative works thereof is used for any 00:13:39.839 --> 00:13:41.910 align:start position:0% derivative works thereof is used for any of<00:13:40.000> your<00:13:40.240> commercial<00:13:40.720> products<00:13:41.120> or<00:13:41.440> services 00:13:41.910 --> 00:13:41.920 align:start position:0% of your commercial products or services 00:13:41.920 --> 00:13:43.750 align:start position:0% of your commercial products or services that<00:13:42.160> have<00:13:42.320> more<00:13:42.560> than<00:13:42.800> 100<00:13:43.040> million<00:13:43.360> monthly 00:13:43.750 --> 00:13:43.760 align:start position:0% that have more than 100 million monthly 00:13:43.760 --> 00:13:46.870 align:start position:0% that have more than 100 million monthly users<00:13:44.320> or<00:13:44.800> more<00:13:45.120> than<00:13:45.440> 2<00:13:45.600> million<00:13:46.000> US<00:13:46.639> or 00:13:46.870 --> 00:13:46.880 align:start position:0% users or more than 2 million US or 00:13:46.880 --> 00:13:48.790 align:start position:0% users or more than 2 million US or equivalent<00:13:47.279> in<00:13:47.440> monthly<00:13:47.760> revenue,<00:13:48.240> you<00:13:48.480> shall 00:13:48.790 --> 00:13:48.800 align:start position:0% equivalent in monthly revenue, you shall 00:13:48.800 --> 00:13:51.430 align:start position:0% equivalent in monthly revenue, you shall prominently<00:13:49.360> display<00:13:49.680> Kimk<00:13:50.320> 2.5<00:13:50.800> on<00:13:51.040> the<00:13:51.200> user 00:13:51.430 --> 00:13:51.440 align:start position:0% prominently display Kimk 2.5 on the user 00:13:51.440 --> 00:13:53.990 align:start position:0% prominently display Kimk 2.5 on the user interface<00:13:52.000> of<00:13:52.240> such<00:13:52.560> product<00:13:53.040> or<00:13:53.440> service. 00:13:53.990 --> 00:13:54.000 align:start position:0% interface of such product or service. 00:13:54.000 --> 00:13:56.790 align:start position:0% interface of such product or service. So,<00:13:54.320> as<00:13:54.560> far<00:13:54.800> as<00:13:55.040> we<00:13:55.360> understand<00:13:56.320> and<00:13:56.560> as<00:13:56.720> far 00:13:56.790 --> 00:13:56.800 align:start position:0% So, as far as we understand and as far 00:13:56.800 --> 00:13:59.189 align:start position:0% So, as far as we understand and as far as<00:13:56.959> I<00:13:57.199> would<00:13:57.360> ever<00:13:57.600> guess,<00:13:58.240> this<00:13:58.560> would<00:13:58.800> be<00:13:58.959> in 00:13:59.189 --> 00:13:59.199 align:start position:0% as I would ever guess, this would be in 00:13:59.199 --> 00:14:01.509 align:start position:0% as I would ever guess, this would be in violation<00:13:59.760> of<00:14:00.000> this<00:14:00.240> license.<00:14:00.959> But<00:14:01.120> nobody's 00:14:01.509 --> 00:14:01.519 align:start position:0% violation of this license. But nobody's 00:14:01.519 --> 00:14:03.110 align:start position:0% violation of this license. But nobody's really<00:14:01.760> tested<00:14:02.079> this<00:14:02.320> yet.<00:14:02.560> And<00:14:02.720> I<00:14:02.880> said<00:14:02.959> this 00:14:03.110 --> 00:14:03.120 align:start position:0% really tested this yet. And I said this 00:14:03.120 --> 00:14:04.629 align:start position:0% really tested this yet. And I said this the<00:14:03.360> last<00:14:03.519> time<00:14:03.600> I<00:14:03.839> covered<00:14:04.160> this<00:14:04.320> type<00:14:04.480> of 00:14:04.629 --> 00:14:04.639 align:start position:0% the last time I covered this type of 00:14:04.639 --> 00:14:06.310 align:start position:0% the last time I covered this type of license<00:14:04.959> in<00:14:05.120> a<00:14:05.279> video<00:14:05.440> and<00:14:05.680> almost<00:14:05.920> every<00:14:06.160> time 00:14:06.310 --> 00:14:06.320 align:start position:0% license in a video and almost every time 00:14:06.320 --> 00:14:08.470 align:start position:0% license in a video and almost every time before<00:14:06.560> it,<00:14:06.800> too.<00:14:07.040> This<00:14:07.279> modification<00:14:08.079> isn't 00:14:08.470 --> 00:14:08.480 align:start position:0% before it, too. This modification isn't 00:14:08.480 --> 00:14:09.590 align:start position:0% before it, too. This modification isn't something<00:14:08.639> that's<00:14:08.880> been<00:14:09.040> tested<00:14:09.279> in<00:14:09.440> the 00:14:09.590 --> 00:14:09.600 align:start position:0% something that's been tested in the 00:14:09.600 --> 00:14:12.470 align:start position:0% something that's been tested in the courts<00:14:09.839> yet.<00:14:10.160> We<00:14:10.399> don't<00:14:10.560> know<00:14:11.519> what<00:14:12.000> this 00:14:12.470 --> 00:14:12.480 align:start position:0% courts yet. We don't know what this 00:14:12.480 --> 00:14:15.189 align:start position:0% courts yet. We don't know what this actually<00:14:13.440> works<00:14:13.760> for.<00:14:14.320> And<00:14:14.480> we<00:14:14.720> also<00:14:14.959> don't 00:14:15.189 --> 00:14:15.199 align:start position:0% actually works for. And we also don't 00:14:15.199 --> 00:14:17.350 align:start position:0% actually works for. And we also don't know<00:14:15.440> how<00:14:15.600> much<00:14:15.760> you<00:14:15.920> can<00:14:16.160> obuscate.<00:14:16.959> If<00:14:17.199> you 00:14:17.350 --> 00:14:17.360 align:start position:0% know how much you can obuscate. If you 00:14:17.360 --> 00:14:19.750 align:start position:0% know how much you can obuscate. If you had<00:14:17.519> a<00:14:17.760> middleman,<00:14:18.399> for<00:14:18.560> example,<00:14:19.279> you<00:14:19.519> made 00:14:19.750 --> 00:14:19.760 align:start position:0% had a middleman, for example, you made 00:14:19.760 --> 00:14:22.790 align:start position:0% had a middleman, for example, you made your<00:14:20.000> own<00:14:20.399> new<00:14:20.639> corporation<00:14:21.680> that<00:14:22.000> thing<00:14:22.320> was 00:14:22.790 --> 00:14:22.800 align:start position:0% your own new corporation that thing was 00:14:22.800 --> 00:14:26.150 align:start position:0% your own new corporation that thing was reselling<00:14:23.360> Chem<00:14:23.760> K2.5<00:14:24.959> and<00:14:25.199> you<00:14:25.440> sold<00:14:25.760> that 00:14:26.150 --> 00:14:26.160 align:start position:0% reselling Chem K2.5 and you sold that 00:14:26.160 --> 00:14:28.870 align:start position:0% reselling Chem K2.5 and you sold that back<00:14:26.480> to<00:14:26.639> your<00:14:26.880> own<00:14:27.040> business<00:14:27.920> for<00:14:28.399> the<00:14:28.639> 100 00:14:28.870 --> 00:14:28.880 align:start position:0% back to your own business for the 100 00:14:28.880 --> 00:14:30.790 align:start position:0% back to your own business for the 100 million<00:14:29.199> monthly<00:14:29.519> users<00:14:29.760> or<00:14:30.000> the<00:14:30.320> 20<00:14:30.480> million 00:14:30.790 --> 00:14:30.800 align:start position:0% million monthly users or the 20 million 00:14:30.800 --> 00:14:32.949 align:start position:0% million monthly users or the 20 million US<00:14:31.279> in<00:14:31.519> revenue.<00:14:31.920> That<00:14:32.160> shell<00:14:32.480> company<00:14:32.720> you 00:14:32.949 --> 00:14:32.959 align:start position:0% US in revenue. That shell company you 00:14:32.959 --> 00:14:34.949 align:start position:0% US in revenue. That shell company you put<00:14:33.120> between<00:14:33.360> the<00:14:33.519> two<00:14:33.839> could<00:14:34.399> be<00:14:34.560> the<00:14:34.800> thing 00:14:34.949 --> 00:14:34.959 align:start position:0% put between the two could be the thing 00:14:34.959 --> 00:14:37.350 align:start position:0% put between the two could be the thing that<00:14:35.199> displays<00:14:35.680> Kimmy<00:14:36.079> K2.5<00:14:36.880> and<00:14:37.040> then<00:14:37.199> it 00:14:37.350 --> 00:14:37.360 align:start position:0% that displays Kimmy K2.5 and then it 00:14:37.360 --> 00:14:39.829 align:start position:0% that displays Kimmy K2.5 and then it sells<00:14:37.680> a<00:14:37.839> service<00:14:38.320> to<00:14:38.560> another<00:14:38.880> company<00:14:39.440> that 00:14:39.829 --> 00:14:39.839 align:start position:0% sells a service to another company that 00:14:39.839 --> 00:14:41.829 align:start position:0% sells a service to another company that doesn't<00:14:40.160> enforce<00:14:40.639> that,<00:14:41.040> that<00:14:41.440> isn't 00:14:41.829 --> 00:14:41.839 align:start position:0% doesn't enforce that, that isn't 00:14:41.839 --> 00:14:43.509 align:start position:0% doesn't enforce that, that isn't affected<00:14:42.399> because<00:14:42.639> they're<00:14:42.959> not<00:14:43.040> agreeing<00:14:43.360> to 00:14:43.509 --> 00:14:43.519 align:start position:0% affected because they're not agreeing to 00:14:43.519 --> 00:14:45.269 align:start position:0% affected because they're not agreeing to this<00:14:43.680> license.<00:14:44.320> The<00:14:44.560> company<00:14:44.800> between<00:14:45.120> the 00:14:45.269 --> 00:14:45.279 align:start position:0% this license. The company between the 00:14:45.279 --> 00:14:47.030 align:start position:0% this license. The company between the two<00:14:45.440> agree<00:14:45.680> to<00:14:45.839> the<00:14:45.920> license.<00:14:46.480> This<00:14:46.720> company's 00:14:47.030 --> 00:14:47.040 align:start position:0% two agree to the license. This company's 00:14:47.040 --> 00:14:48.629 align:start position:0% two agree to the license. This company's just<00:14:47.199> buying<00:14:47.440> a<00:14:47.600> service.<00:14:48.160> This<00:14:48.320> is<00:14:48.480> an 00:14:48.629 --> 00:14:48.639 align:start position:0% just buying a service. This is an 00:14:48.639 --> 00:14:50.069 align:start position:0% just buying a service. This is an important<00:14:48.880> distinction<00:14:49.360> for<00:14:49.600> reasons<00:14:49.839> that 00:14:50.069 --> 00:14:50.079 align:start position:0% important distinction for reasons that 00:14:50.079 --> 00:14:51.990 align:start position:0% important distinction for reasons that you<00:14:50.240> will<00:14:50.560> learn<00:14:50.880> momentarily.<00:14:51.760> As 00:14:51.990 --> 00:14:52.000 align:start position:0% you will learn momentarily. As 00:14:52.000 --> 00:14:54.550 align:start position:0% you will learn momentarily. As speculation<00:14:52.560> caught<00:14:52.959> fire<00:14:53.440> on<00:14:53.680> Twitter,<00:14:54.240> Lee 00:14:54.550 --> 00:14:54.560 align:start position:0% speculation caught fire on Twitter, Lee 00:14:54.560 --> 00:14:56.790 align:start position:0% speculation caught fire on Twitter, Lee Rob,<00:14:55.120> everyone's<00:14:55.600> favorite,<00:14:56.079> came<00:14:56.320> in<00:14:56.480> out<00:14:56.639> of 00:14:56.790 --> 00:14:56.800 align:start position:0% Rob, everyone's favorite, came in out of 00:14:56.800 --> 00:14:57.990 align:start position:0% Rob, everyone's favorite, came in out of nowhere<00:14:57.120> and<00:14:57.279> actually<00:14:57.440> has<00:14:57.680> some<00:14:57.839> really 00:14:57.990 --> 00:14:58.000 align:start position:0% nowhere and actually has some really 00:14:58.000 --> 00:14:59.829 align:start position:0% nowhere and actually has some really good<00:14:58.160> things<00:14:58.240> to<00:14:58.480> say.<00:14:58.880> Composer<00:14:59.360> 2<00:14:59.600> started 00:14:59.829 --> 00:14:59.839 align:start position:0% good things to say. Composer 2 started 00:14:59.839 --> 00:15:01.910 align:start position:0% good things to say. Composer 2 started from<00:15:00.000> an<00:15:00.240> open-<00:15:00.560> source<00:15:00.880> base.<00:15:01.360> We<00:15:01.600> will<00:15:01.760> do 00:15:01.910 --> 00:15:01.920 align:start position:0% from an open- source base. We will do 00:15:01.920 --> 00:15:04.069 align:start position:0% from an open- source base. We will do full<00:15:02.240> pre-training<00:15:02.880> in<00:15:03.120> the<00:15:03.279> future.<00:15:03.839> As<00:15:04.000> I 00:15:04.069 --> 00:15:04.079 align:start position:0% full pre-training in the future. As I 00:15:04.079 --> 00:15:05.750 align:start position:0% full pre-training in the future. As I mentioned<00:15:04.399> before,<00:15:04.639> the<00:15:04.800> pre-training<00:15:05.440> is 00:15:05.750 --> 00:15:05.760 align:start position:0% mentioned before, the pre-training is 00:15:05.760 --> 00:15:07.350 align:start position:0% mentioned before, the pre-training is how<00:15:05.920> you<00:15:06.079> get<00:15:06.240> a<00:15:06.399> lot<00:15:06.480> of<00:15:06.560> the<00:15:06.720> knowledge<00:15:07.040> into 00:15:07.350 --> 00:15:07.360 align:start position:0% how you get a lot of the knowledge into 00:15:07.360 --> 00:15:08.710 align:start position:0% how you get a lot of the knowledge into the<00:15:07.519> model.<00:15:07.760> It's<00:15:08.000> the<00:15:08.240> process<00:15:08.399> where<00:15:08.560> you 00:15:08.710 --> 00:15:08.720 align:start position:0% the model. It's the process where you 00:15:08.720 --> 00:15:11.509 align:start position:0% the model. It's the process where you take<00:15:08.959> a<00:15:09.199> shitload<00:15:09.600> of<00:15:09.760> data<00:15:10.320> and<00:15:10.720> do<00:15:10.959> a<00:15:11.279> bunch 00:15:11.509 --> 00:15:11.519 align:start position:0% take a shitload of data and do a bunch 00:15:11.519 --> 00:15:13.350 align:start position:0% take a shitload of data and do a bunch of<00:15:11.839> transformations<00:15:12.399> on<00:15:12.639> it<00:15:12.720> to<00:15:12.880> try<00:15:13.040> and<00:15:13.199> make 00:15:13.350 --> 00:15:13.360 align:start position:0% of transformations on it to try and make 00:15:13.360 --> 00:15:16.069 align:start position:0% of transformations on it to try and make the<00:15:13.519> set<00:15:13.680> of<00:15:14.000> parameters<00:15:14.639> that<00:15:14.959> represent<00:15:15.839> the 00:15:16.069 --> 00:15:16.079 align:start position:0% the set of parameters that represent the 00:15:16.079 --> 00:15:18.230 align:start position:0% the set of parameters that represent the knowledge<00:15:16.399> that<00:15:16.639> it<00:15:16.800> is<00:15:16.959> capable<00:15:17.279> of.<00:15:17.680> RL<00:15:18.079> is 00:15:18.230 --> 00:15:18.240 align:start position:0% knowledge that it is capable of. RL is 00:15:18.240 --> 00:15:20.710 align:start position:0% knowledge that it is capable of. RL is mostly<00:15:18.560> a<00:15:19.040> post-training<00:15:19.920> process,<00:15:20.480> but 00:15:20.710 --> 00:15:20.720 align:start position:0% mostly a post-training process, but 00:15:20.720 --> 00:15:22.629 align:start position:0% mostly a post-training process, but what's<00:15:21.040> really<00:15:21.279> interesting<00:15:21.680> here<00:15:22.000> is<00:15:22.320> the 00:15:22.629 --> 00:15:22.639 align:start position:0% what's really interesting here is the 00:15:22.639 --> 00:15:24.790 align:start position:0% what's really interesting here is the compute<00:15:23.040> split.<00:15:23.360> Kimmy<00:15:23.680> K2.5<00:15:24.320> is<00:15:24.480> a<00:15:24.639> great 00:15:24.790 --> 00:15:24.800 align:start position:0% compute split. Kimmy K2.5 is a great 00:15:24.800 --> 00:15:26.389 align:start position:0% compute split. Kimmy K2.5 is a great model.<00:15:25.120> As<00:15:25.360> someone<00:15:25.519> who's<00:15:25.839> been<00:15:25.920> a<00:15:26.079> fan<00:15:26.240> of 00:15:26.389 --> 00:15:26.399 align:start position:0% model. As someone who's been a fan of 00:15:26.399 --> 00:15:28.069 align:start position:0% model. As someone who's been a fan of Moonshot<00:15:26.800> and<00:15:26.959> Kimmy<00:15:27.279> for<00:15:27.360> a<00:15:27.519> while,<00:15:27.760> this<00:15:28.000> was 00:15:28.069 --> 00:15:28.079 align:start position:0% Moonshot and Kimmy for a while, this was 00:15:28.079 --> 00:15:30.470 align:start position:0% Moonshot and Kimmy for a while, this was a<00:15:28.399> very<00:15:28.720> fun<00:15:29.040> week<00:15:29.279> for<00:15:29.440> me<00:15:29.839> because<00:15:30.079> I've<00:15:30.320> been 00:15:30.470 --> 00:15:30.480 align:start position:0% a very fun week for me because I've been 00:15:30.480 --> 00:15:31.829 align:start position:0% a very fun week for me because I've been saying<00:15:30.639> for<00:15:30.720> a<00:15:30.880> bit<00:15:30.959> that<00:15:31.120> the<00:15:31.279> Kimmy<00:15:31.600> models 00:15:31.829 --> 00:15:31.839 align:start position:0% saying for a bit that the Kimmy models 00:15:31.839 --> 00:15:33.590 align:start position:0% saying for a bit that the Kimmy models are<00:15:32.000> underrated<00:15:32.480> and<00:15:32.720> really<00:15:32.959> good.<00:15:33.279> To<00:15:33.440> this 00:15:33.590 --> 00:15:33.600 align:start position:0% are underrated and really good. To this 00:15:33.600 --> 00:15:35.910 align:start position:0% are underrated and really good. To this day,<00:15:33.839> Kimmy<00:15:34.160> K2<00:15:34.639> is<00:15:34.720> the<00:15:34.880> default<00:15:35.120> model<00:15:35.360> in<00:15:35.519> T3 00:15:35.910 --> 00:15:35.920 align:start position:0% day, Kimmy K2 is the default model in T3 00:15:35.920 --> 00:15:37.189 align:start position:0% day, Kimmy K2 is the default model in T3 checks.<00:15:36.160> I<00:15:36.240> think<00:15:36.399> it's<00:15:36.560> one<00:15:36.639> of<00:15:36.720> the<00:15:36.880> best 00:15:37.189 --> 00:15:37.199 align:start position:0% checks. I think it's one of the best 00:15:37.199 --> 00:15:39.750 align:start position:0% checks. I think it's one of the best options<00:15:37.839> for<00:15:38.240> just<00:15:38.560> talking<00:15:38.880> to<00:15:39.120> a<00:15:39.279> model.<00:15:39.519> Its 00:15:39.750 --> 00:15:39.760 align:start position:0% options for just talking to a model. Its 00:15:39.760 --> 00:15:41.430 align:start position:0% options for just talking to a model. Its pros<00:15:40.079> is<00:15:40.320> great.<00:15:40.560> Its<00:15:40.800> knowledge<00:15:41.120> is<00:15:41.279> more 00:15:41.430 --> 00:15:41.440 align:start position:0% pros is great. Its knowledge is more 00:15:41.440 --> 00:15:43.269 align:start position:0% pros is great. Its knowledge is more than<00:15:41.600> good<00:15:41.760> enough.<00:15:42.160> It's<00:15:42.480> fast.<00:15:42.880> It's<00:15:43.040> nice. 00:15:43.269 --> 00:15:43.279 align:start position:0% than good enough. It's fast. It's nice. 00:15:43.279 --> 00:15:45.189 align:start position:0% than good enough. It's fast. It's nice. It's<00:15:43.519> just<00:15:43.600> a<00:15:43.839> good<00:15:44.079> model.<00:15:44.560> I<00:15:44.800> genuinely 00:15:45.189 --> 00:15:45.199 align:start position:0% It's just a good model. I genuinely 00:15:45.199 --> 00:15:47.829 align:start position:0% It's just a good model. I genuinely enjoy<00:15:45.519> using<00:15:45.760> it<00:15:45.920> a<00:15:46.079> lot.<00:15:46.399> And<00:15:46.560> 2.5<00:15:47.199> was<00:15:47.519> much 00:15:47.829 --> 00:15:47.839 align:start position:0% enjoy using it a lot. And 2.5 was much 00:15:47.839 --> 00:15:49.590 align:start position:0% enjoy using it a lot. And 2.5 was much smarter<00:15:48.160> and<00:15:48.320> more<00:15:48.560> capable.<00:15:49.040> Slightly<00:15:49.360> worse 00:15:49.590 --> 00:15:49.600 align:start position:0% smarter and more capable. Slightly worse 00:15:49.600 --> 00:15:51.189 align:start position:0% smarter and more capable. Slightly worse at<00:15:49.759> writing,<00:15:50.079> but<00:15:50.480> still<00:15:50.720> really<00:15:50.959> good 00:15:51.189 --> 00:15:51.199 align:start position:0% at writing, but still really good 00:15:51.199 --> 00:15:53.430 align:start position:0% at writing, but still really good overall.<00:15:51.759> Turns<00:15:52.000> out<00:15:52.240> Cursor<00:15:52.800> agreed<00:15:53.120> and 00:15:53.430 --> 00:15:53.440 align:start position:0% overall. Turns out Cursor agreed and 00:15:53.440 --> 00:15:55.829 align:start position:0% overall. Turns out Cursor agreed and chose<00:15:53.680> Kimmy<00:15:54.079> Kate<00:15:54.320> 2.5<00:15:54.800> as<00:15:55.040> the<00:15:55.279> base<00:15:55.519> model 00:15:55.829 --> 00:15:55.839 align:start position:0% chose Kimmy Kate 2.5 as the base model 00:15:55.839 --> 00:15:57.590 align:start position:0% chose Kimmy Kate 2.5 as the base model that<00:15:56.079> they<00:15:56.240> would<00:15:56.399> do<00:15:56.480> all<00:15:56.720> of<00:15:56.800> their<00:15:56.959> RL<00:15:57.360> on 00:15:57.590 --> 00:15:57.600 align:start position:0% that they would do all of their RL on 00:15:57.600 --> 00:15:59.350 align:start position:0% that they would do all of their RL on using<00:15:57.839> their<00:15:58.160> insane<00:15:58.560> amount<00:15:58.720> of<00:15:58.880> data,<00:15:59.199> all 00:15:59.350 --> 00:15:59.360 align:start position:0% using their insane amount of data, all 00:15:59.360 --> 00:16:00.550 align:start position:0% using their insane amount of data, all of<00:15:59.440> their<00:15:59.519> harnesses,<00:16:00.079> all<00:16:00.160> of<00:16:00.240> the<00:16:00.399> other 00:16:00.550 --> 00:16:00.560 align:start position:0% of their harnesses, all of the other 00:16:00.560 --> 00:16:02.069 align:start position:0% of their harnesses, all of the other stuff<00:16:00.720> they've<00:16:01.040> built<00:16:01.360> with<00:16:01.519> their<00:16:01.839> small 00:16:02.069 --> 00:16:02.079 align:start position:0% stuff they've built with their small 00:16:02.079 --> 00:16:04.230 align:start position:0% stuff they've built with their small team<00:16:02.320> of<00:16:02.480> researchers<00:16:03.199> to<00:16:03.519> try<00:16:03.680> and<00:16:03.920> make 00:16:04.230 --> 00:16:04.240 align:start position:0% team of researchers to try and make 00:16:04.240 --> 00:16:06.949 align:start position:0% team of researchers to try and make Kimmy<00:16:04.800> the<00:16:05.040> best<00:16:05.360> coding<00:16:05.839> model<00:16:06.240> per<00:16:06.560> price. 00:16:06.949 --> 00:16:06.959 align:start position:0% Kimmy the best coding model per price. 00:16:06.959 --> 00:16:08.069 align:start position:0% Kimmy the best coding model per price. And<00:16:07.040> that's<00:16:07.199> why<00:16:07.360> they<00:16:07.519> chose<00:16:07.600> to<00:16:07.680> use<00:16:07.839> Kimmy 00:16:08.069 --> 00:16:08.079 align:start position:0% And that's why they chose to use Kimmy 00:16:08.079 --> 00:16:09.990 align:start position:0% And that's why they chose to use Kimmy K2.5<00:16:08.720> as<00:16:08.880> the<00:16:09.040> starting<00:16:09.360> point<00:16:09.519> for<00:16:09.759> this 00:16:09.990 --> 00:16:10.000 align:start position:0% K2.5 as the starting point for this 00:16:10.000 --> 00:16:12.790 align:start position:0% K2.5 as the starting point for this crazy<00:16:10.720> pile<00:16:11.120> of<00:16:11.360> post-<00:16:11.680> training<00:16:11.920> in<00:16:12.079> RL<00:16:12.399> on<00:16:12.560> an 00:16:12.790 --> 00:16:12.800 align:start position:0% crazy pile of post- training in RL on an 00:16:12.800 --> 00:16:14.310 align:start position:0% crazy pile of post- training in RL on an existing<00:16:13.040> model<00:16:13.519> so<00:16:13.680> they<00:16:13.839> could<00:16:14.000> try<00:16:14.160> and 00:16:14.310 --> 00:16:14.320 align:start position:0% existing model so they could try and 00:16:14.320 --> 00:16:16.150 align:start position:0% existing model so they could try and make<00:16:14.480> it<00:16:14.639> the<00:16:14.880> best<00:16:15.120> coding<00:16:15.519> model<00:16:15.759> for<00:16:16.000> the 00:16:16.150 --> 00:16:16.160 align:start position:0% make it the best coding model for the 00:16:16.160 --> 00:16:18.069 align:start position:0% make it the best coding model for the price<00:16:16.720> ever.<00:16:17.279> They<00:16:17.440> wanted<00:16:17.600> a<00:16:17.759> really<00:16:17.920> good 00:16:18.069 --> 00:16:18.079 align:start position:0% price ever. They wanted a really good 00:16:18.079 --> 00:16:19.509 align:start position:0% price ever. They wanted a really good coding<00:16:18.399> model<00:16:18.560> that<00:16:18.720> worked<00:16:18.959> great<00:16:19.120> in<00:16:19.360> their 00:16:19.509 --> 00:16:19.519 align:start position:0% coding model that worked great in their 00:16:19.519 --> 00:16:21.430 align:start position:0% coding model that worked great in their harness<00:16:20.000> that<00:16:20.240> they<00:16:20.480> could<00:16:20.639> serve<00:16:20.880> for<00:16:21.120> cheap 00:16:21.430 --> 00:16:21.440 align:start position:0% harness that they could serve for cheap 00:16:21.440 --> 00:16:23.749 align:start position:0% harness that they could serve for cheap and<00:16:21.680> really<00:16:21.920> fast.<00:16:22.399> And<00:16:22.560> they<00:16:22.800> went<00:16:23.120> well<00:16:23.440> past 00:16:23.749 --> 00:16:23.759 align:start position:0% and really fast. And they went well past 00:16:23.759 --> 00:16:25.189 align:start position:0% and really fast. And they went well past what<00:16:24.000> I<00:16:24.160> would<00:16:24.320> imagine<00:16:24.639> even<00:16:24.880> they<00:16:25.120> thought 00:16:25.189 --> 00:16:25.199 align:start position:0% what I would imagine even they thought 00:16:25.199 --> 00:16:26.870 align:start position:0% what I would imagine even they thought they'd<00:16:25.440> be<00:16:25.519> able<00:16:25.680> to<00:16:25.759> do<00:16:25.920> with<00:16:26.079> this<00:16:26.240> strategy. 00:16:26.870 --> 00:16:26.880 align:start position:0% they'd be able to do with this strategy. 00:16:26.880 --> 00:16:28.470 align:start position:0% they'd be able to do with this strategy. Taking<00:16:27.120> this<00:16:27.360> model<00:16:27.600> that<00:16:27.839> was<00:16:28.000> pretty<00:16:28.320> good 00:16:28.470 --> 00:16:28.480 align:start position:0% Taking this model that was pretty good 00:16:28.480 --> 00:16:31.670 align:start position:0% Taking this model that was pretty good before<00:16:29.120> and<00:16:29.279> bumping<00:16:29.680> it<00:16:29.839> to<00:16:30.079> be<00:16:30.240> in<00:16:30.480> the<00:16:30.639> GPT5 00:16:31.670 --> 00:16:31.680 align:start position:0% before and bumping it to be in the GPT5 00:16:31.680 --> 00:16:33.430 align:start position:0% before and bumping it to be in the GPT5 tier<00:16:32.079> of<00:16:32.240> capability<00:16:32.880> according<00:16:33.199> to<00:16:33.279> their 00:16:33.430 --> 00:16:33.440 align:start position:0% tier of capability according to their 00:16:33.440 --> 00:16:35.829 align:start position:0% tier of capability according to their own<00:16:33.600> benchmarks<00:16:34.320> is<00:16:34.639> nuts.<00:16:35.199> They<00:16:35.360> did<00:16:35.519> run<00:16:35.680> it 00:16:35.829 --> 00:16:35.839 align:start position:0% own benchmarks is nuts. They did run it 00:16:35.839 --> 00:16:37.910 align:start position:0% own benchmarks is nuts. They did run it on<00:16:36.079> other<00:16:36.399> external<00:16:37.120> code<00:16:37.360> benches<00:16:37.759> like 00:16:37.910 --> 00:16:37.920 align:start position:0% on other external code benches like 00:16:37.920 --> 00:16:39.590 align:start position:0% on other external code benches like terminal<00:16:38.320> bench<00:16:38.639> 2<00:16:38.880> where<00:16:39.120> it<00:16:39.360> still 00:16:39.590 --> 00:16:39.600 align:start position:0% terminal bench 2 where it still 00:16:39.600 --> 00:16:42.710 align:start position:0% terminal bench 2 where it still outscored<00:16:40.160> Opus<00:16:40.639> 45<00:16:40.959> and<00:16:41.199> 46.<00:16:42.160> Seems<00:16:42.399> to<00:16:42.560> be<00:16:42.639> a 00:16:42.710 --> 00:16:42.720 align:start position:0% outscored Opus 45 and 46. Seems to be a 00:16:42.720 --> 00:16:44.069 align:start position:0% outscored Opus 45 and 46. Seems to be a good<00:16:42.880> code<00:16:43.040> model<00:16:43.279> and<00:16:43.440> having<00:16:43.680> used<00:16:43.839> it<00:16:43.920> a 00:16:44.069 --> 00:16:44.079 align:start position:0% good code model and having used it a 00:16:44.079 --> 00:16:45.910 align:start position:0% good code model and having used it a bunch<00:16:44.320> myself,<00:16:44.720> it's<00:16:45.040> a<00:16:45.279> pretty<00:16:45.519> dang<00:16:45.759> good 00:16:45.910 --> 00:16:45.920 align:start position:0% bunch myself, it's a pretty dang good 00:16:45.920 --> 00:16:47.509 align:start position:0% bunch myself, it's a pretty dang good model.<00:16:46.320> It<00:16:46.560> makes<00:16:46.720> dumb<00:16:46.959> mistakes<00:16:47.199> here<00:16:47.360> and 00:16:47.509 --> 00:16:47.519 align:start position:0% model. It makes dumb mistakes here and 00:16:47.519 --> 00:16:48.550 align:start position:0% model. It makes dumb mistakes here and there.<00:16:47.680> You<00:16:47.759> have<00:16:47.839> to<00:16:47.920> steer<00:16:48.160> it<00:16:48.240> every<00:16:48.399> once 00:16:48.550 --> 00:16:48.560 align:start position:0% there. You have to steer it every once 00:16:48.560 --> 00:16:50.550 align:start position:0% there. You have to steer it every once in<00:16:48.720> a<00:16:48.800> while,<00:16:49.279> but<00:16:49.440> it's<00:16:49.680> so<00:16:49.920> cheap<00:16:50.160> and<00:16:50.399> so 00:16:50.550 --> 00:16:50.560 align:start position:0% in a while, but it's so cheap and so 00:16:50.560 --> 00:16:52.870 align:start position:0% in a while, but it's so cheap and so fast<00:16:51.040> that<00:16:51.279> I<00:16:51.440> find<00:16:51.600> it<00:16:51.759> really<00:16:52.000> nice<00:16:52.160> to<00:16:52.399> use. 00:16:52.870 --> 00:16:52.880 align:start position:0% fast that I find it really nice to use. 00:16:52.880 --> 00:16:54.389 align:start position:0% fast that I find it really nice to use. It's<00:16:53.120> also<00:16:53.279> pretty<00:16:53.519> good<00:16:53.600> at<00:16:53.839> design.<00:16:54.160> Shout 00:16:54.389 --> 00:16:54.399 align:start position:0% It's also pretty good at design. Shout 00:16:54.399 --> 00:16:56.629 align:start position:0% It's also pretty good at design. Shout out<00:16:54.480> to<00:16:54.639> D,<00:16:55.199> by<00:16:55.279> the<00:16:55.440> way,<00:16:55.519> for<00:16:55.839> making<00:16:56.079> this 00:16:56.629 --> 00:16:56.639 align:start position:0% out to D, by the way, for making this 00:16:56.639 --> 00:16:59.030 align:start position:0% out to D, by the way, for making this very<00:16:57.040> useful<00:16:57.440> website<00:16:57.920> where<00:16:58.079> I<00:16:58.320> can<00:16:58.480> compare 00:16:59.030 --> 00:16:59.040 align:start position:0% very useful website where I can compare 00:16:59.040 --> 00:17:01.110 align:start position:0% very useful website where I can compare runs<00:16:59.360> on<00:16:59.600> a<00:16:59.759> specific<00:17:00.160> prompt<00:17:00.639> with<00:17:00.880> and 00:17:01.110 --> 00:17:01.120 align:start position:0% runs on a specific prompt with and 00:17:01.120 --> 00:17:02.870 align:start position:0% runs on a specific prompt with and without<00:17:01.440> the<00:17:01.759> anthropic<00:17:02.240> design<00:17:02.480> skill 00:17:02.870 --> 00:17:02.880 align:start position:0% without the anthropic design skill 00:17:02.880 --> 00:17:04.949 align:start position:0% without the anthropic design skill across<00:17:03.360> various<00:17:03.759> models.<00:17:04.319> I<00:17:04.559> think<00:17:04.640> this<00:17:04.799> will 00:17:04.949 --> 00:17:04.959 align:start position:0% across various models. I think this will 00:17:04.959 --> 00:17:06.789 align:start position:0% across various models. I think this will be<00:17:05.039> fun.<00:17:05.360> Here's<00:17:05.679> an<00:17:05.760> example<00:17:06.079> of<00:17:06.240> a<00:17:06.400> design 00:17:06.789 --> 00:17:06.799 align:start position:0% be fun. Here's an example of a design 00:17:06.799 --> 00:17:09.909 align:start position:0% be fun. Here's an example of a design that<00:17:07.039> was<00:17:07.199> made<00:17:07.439> with<00:17:07.679> Composer<00:17:08.319> 2.0<00:17:09.120> for<00:17:09.439> this 00:17:09.909 --> 00:17:09.919 align:start position:0% that was made with Composer 2.0 for this 00:17:09.919 --> 00:17:11.590 align:start position:0% that was made with Composer 2.0 for this example<00:17:10.400> site.<00:17:10.720> I<00:17:10.959> think<00:17:11.039> this<00:17:11.199> looks<00:17:11.439> quite 00:17:11.590 --> 00:17:11.600 align:start position:0% example site. I think this looks quite 00:17:11.600 --> 00:17:13.510 align:start position:0% example site. I think this looks quite good.<00:17:11.919> And<00:17:12.000> we<00:17:12.160> look<00:17:12.319> through<00:17:12.480> these.<00:17:12.880> Yeah,<00:17:13.280> I 00:17:13.510 --> 00:17:13.520 align:start position:0% good. And we look through these. Yeah, I 00:17:13.520 --> 00:17:17.270 align:start position:0% good. And we look through these. Yeah, I hate<00:17:13.760> these<00:17:14.240> retro<00:17:15.360> blue<00:17:15.760> pink<00:17:16.400> glowy<00:17:16.799> styles, 00:17:17.270 --> 00:17:17.280 align:start position:0% hate these retro blue pink glowy styles, 00:17:17.280 --> 00:17:18.949 align:start position:0% hate these retro blue pink glowy styles, but<00:17:17.439> the<00:17:17.679> other<00:17:17.760> ones<00:17:18.000> all<00:17:18.240> look<00:17:18.640> really 00:17:18.949 --> 00:17:18.959 align:start position:0% but the other ones all look really 00:17:18.959 --> 00:17:20.549 align:start position:0% but the other ones all look really solid.<00:17:19.439> These<00:17:19.679> aren't<00:17:19.919> bad.<00:17:20.079> And<00:17:20.240> if<00:17:20.400> we 00:17:20.549 --> 00:17:20.559 align:start position:0% solid. These aren't bad. And if we 00:17:20.559 --> 00:17:22.630 align:start position:0% solid. These aren't bad. And if we compare<00:17:20.880> to<00:17:21.120> Kimmy<00:17:21.439> K2.5,<00:17:22.400> there's 00:17:22.630 --> 00:17:22.640 align:start position:0% compare to Kimmy K2.5, there's 00:17:22.640 --> 00:17:24.309 align:start position:0% compare to Kimmy K2.5, there's definitely<00:17:23.120> overlap<00:17:23.520> in<00:17:23.760> some<00:17:23.919> of<00:17:24.000> these 00:17:24.309 --> 00:17:24.319 align:start position:0% definitely overlap in some of these 00:17:24.319 --> 00:17:27.189 align:start position:0% definitely overlap in some of these designs,<00:17:25.039> but<00:17:25.280> overall<00:17:26.240> the<00:17:26.400> composer<00:17:26.880> ones 00:17:27.189 --> 00:17:27.199 align:start position:0% designs, but overall the composer ones 00:17:27.199 --> 00:17:29.110 align:start position:0% designs, but overall the composer ones are<00:17:27.439> slightly<00:17:27.839> better.<00:17:28.319> Not<00:17:28.480> a<00:17:28.720> lot<00:17:28.799> better, 00:17:29.110 --> 00:17:29.120 align:start position:0% are slightly better. Not a lot better, 00:17:29.120 --> 00:17:30.710 align:start position:0% are slightly better. Not a lot better, but<00:17:29.360> enough<00:17:29.679> better<00:17:29.840> that<00:17:30.080> it's<00:17:30.240> pretty<00:17:30.480> cool. 00:17:30.710 --> 00:17:30.720 align:start position:0% but enough better that it's pretty cool. 00:17:30.720 --> 00:17:32.470 align:start position:0% but enough better that it's pretty cool. I<00:17:30.960> still<00:17:31.039> think<00:17:31.280> Opus<00:17:31.600> 4.6<00:17:32.080> with<00:17:32.160> the<00:17:32.320> front 00:17:32.470 --> 00:17:32.480 align:start position:0% I still think Opus 4.6 with the front 00:17:32.480 --> 00:17:34.470 align:start position:0% I still think Opus 4.6 with the front end<00:17:32.640> design<00:17:32.880> skill<00:17:33.200> is<00:17:33.440> the<00:17:33.600> most<00:17:33.919> reliable 00:17:34.470 --> 00:17:34.480 align:start position:0% end design skill is the most reliable 00:17:34.480 --> 00:17:36.789 align:start position:0% end design skill is the most reliable for<00:17:34.720> good<00:17:34.960> designs<00:17:35.440> overall.<00:17:36.160> But<00:17:36.320> it's 00:17:36.789 --> 00:17:36.799 align:start position:0% for good designs overall. But it's 00:17:36.799 --> 00:17:38.150 align:start position:0% for good designs overall. But it's actually<00:17:37.039> the<00:17:37.200> gap's<00:17:37.520> not<00:17:37.600> that<00:17:37.760> big<00:17:37.919> here.<00:17:38.080> I 00:17:38.150 --> 00:17:38.160 align:start position:0% actually the gap's not that big here. I 00:17:38.160 --> 00:17:40.710 align:start position:0% actually the gap's not that big here. I think<00:17:38.320> Compose<00:17:38.640> are<00:17:38.799> kind<00:17:38.960> of<00:17:39.039> cooked.<00:17:39.919> Yeah, 00:17:40.710 --> 00:17:40.720 align:start position:0% think Compose are kind of cooked. Yeah, 00:17:40.720 --> 00:17:42.870 align:start position:0% think Compose are kind of cooked. Yeah, good<00:17:41.760> Yeah,<00:17:42.000> I'm<00:17:42.160> impressed.<00:17:42.480> Composer 00:17:42.870 --> 00:17:42.880 align:start position:0% good Yeah, I'm impressed. Composer 00:17:42.880 --> 00:17:44.549 align:start position:0% good Yeah, I'm impressed. Composer did<00:17:43.039> very<00:17:43.200> well<00:17:43.360> here.<00:17:43.679> It's<00:17:43.840> a<00:17:44.000> good<00:17:44.160> model. 00:17:44.549 --> 00:17:44.559 align:start position:0% did very well here. It's a good model. 00:17:44.559 --> 00:17:47.750 align:start position:0% did very well here. It's a good model. It's<00:17:44.880> very<00:17:45.280> solid.<00:17:46.080> And<00:17:46.240> the<00:17:46.559> speed<00:17:46.880> in<00:17:47.200> price 00:17:47.750 --> 00:17:47.760 align:start position:0% It's very solid. And the speed in price 00:17:47.760 --> 00:17:50.470 align:start position:0% It's very solid. And the speed in price is<00:17:48.080> insane.<00:17:48.640> Their<00:17:48.960> default<00:17:49.360> version<00:17:49.760> does<00:17:50.320> 80 00:17:50.470 --> 00:17:50.480 align:start position:0% is insane. Their default version does 80 00:17:50.480 --> 00:17:53.430 align:start position:0% is insane. Their default version does 80 to<00:17:50.720> 100<00:17:51.039> TPS<00:17:51.679> and<00:17:51.919> it<00:17:52.160> costs<00:17:52.799> basically 00:17:53.430 --> 00:17:53.440 align:start position:0% to 100 TPS and it costs basically 00:17:53.440 --> 00:17:55.909 align:start position:0% to 100 TPS and it costs basically nothing.<00:17:54.160> Yeah,<00:17:54.640> 50<00:17:54.799> cents<00:17:55.039> per<00:17:55.280> million<00:17:55.600> in 00:17:55.909 --> 00:17:55.919 align:start position:0% nothing. Yeah, 50 cents per million in 00:17:55.919 --> 00:17:58.870 align:start position:0% nothing. Yeah, 50 cents per million in and<00:17:56.320> 250<00:17:56.720> per<00:17:56.960> million<00:17:57.280> out.<00:17:57.760> For<00:17:58.000> comparison, 00:17:58.870 --> 00:17:58.880 align:start position:0% and 250 per million out. For comparison, 00:17:58.880 --> 00:18:02.390 align:start position:0% and 250 per million out. For comparison, Opus<00:17:59.360> is<00:17:59.840> $5<00:18:00.320> per<00:18:00.480> mill<00:18:00.720> in<00:18:00.880> and<00:18:01.200> $25<00:18:01.919> per<00:18:02.160> mill 00:18:02.390 --> 00:18:02.400 align:start position:0% Opus is $5 per mill in and $25 per mill 00:18:02.400 --> 00:18:05.590 align:start position:0% Opus is $5 per mill in and $25 per mill out.<00:18:03.039> Yes,<00:18:03.440> that<00:18:03.679> means<00:18:03.840> it's<00:18:04.160> exactly<00:18:05.039> 10x 00:18:05.590 --> 00:18:05.600 align:start position:0% out. Yes, that means it's exactly 10x 00:18:05.600 --> 00:18:07.909 align:start position:0% out. Yes, that means it's exactly 10x cheaper<00:18:06.000> than<00:18:06.160> Opus<00:18:06.640> 46.<00:18:07.200> That<00:18:07.360> is<00:18:07.520> insane, 00:18:07.909 --> 00:18:07.919 align:start position:0% cheaper than Opus 46. That is insane, 00:18:07.919 --> 00:18:09.270 align:start position:0% cheaper than Opus 46. That is insane, but<00:18:08.080> that<00:18:08.240> doesn't<00:18:08.480> mean<00:18:08.559> they<00:18:08.799> can<00:18:08.880> just<00:18:09.120> get 00:18:09.270 --> 00:18:09.280 align:start position:0% but that doesn't mean they can just get 00:18:09.280 --> 00:18:12.070 align:start position:0% but that doesn't mean they can just get away<00:18:09.520> with<00:18:09.679> this.<00:18:09.919> So<00:18:10.160> back<00:18:10.480> to<00:18:11.280> these<00:18:11.520> post 00:18:12.070 --> 00:18:12.080 align:start position:0% away with this. So back to these post 00:18:12.080 --> 00:18:13.830 align:start position:0% away with this. So back to these post turns<00:18:12.320> out<00:18:12.480> that<00:18:12.720> post<00:18:12.960> training<00:18:13.360> is<00:18:13.600> very 00:18:13.830 --> 00:18:13.840 align:start position:0% turns out that post training is very 00:18:13.840 --> 00:18:15.669 align:start position:0% turns out that post training is very valuable<00:18:14.160> because<00:18:14.400> Kimmy<00:18:14.799> K25<00:18:15.280> didn't<00:18:15.520> get 00:18:15.669 --> 00:18:15.679 align:start position:0% valuable because Kimmy K25 didn't get 00:18:15.679 --> 00:18:17.350 align:start position:0% valuable because Kimmy K25 didn't get close<00:18:15.840> to<00:18:16.000> any<00:18:16.160> of<00:18:16.240> these<00:18:16.480> numbers<00:18:16.960> and<00:18:17.120> that's 00:18:17.350 --> 00:18:17.360 align:start position:0% close to any of these numbers and that's 00:18:17.360 --> 00:18:19.350 align:start position:0% close to any of these numbers and that's because<00:18:17.600> Kimmy<00:18:18.000> K25<00:18:18.720> was<00:18:19.039> only 00:18:19.350 --> 00:18:19.360 align:start position:0% because Kimmy K25 was only 00:18:19.360 --> 00:18:20.710 align:start position:0% because Kimmy K25 was only representative<00:18:19.840> of<00:18:19.919> a<00:18:20.160> fourth<00:18:20.400> of<00:18:20.559> the 00:18:20.710 --> 00:18:20.720 align:start position:0% representative of a fourth of the 00:18:20.720 --> 00:18:22.470 align:start position:0% representative of a fourth of the compute<00:18:21.120> they<00:18:21.360> used<00:18:21.600> for<00:18:21.760> the<00:18:21.919> training<00:18:22.160> here. 00:18:22.470 --> 00:18:22.480 align:start position:0% compute they used for the training here. 00:18:22.480 --> 00:18:25.029 align:start position:0% compute they used for the training here. K25<00:18:23.280> is<00:18:23.520> the<00:18:23.760> model<00:18:24.000> we<00:18:24.240> all<00:18:24.480> have<00:18:24.640> access<00:18:24.880> to 00:18:25.029 --> 00:18:25.039 align:start position:0% K25 is the model we all have access to 00:18:25.039 --> 00:18:26.789 align:start position:0% K25 is the model we all have access to to<00:18:25.280> download<00:18:25.520> and<00:18:25.679> use<00:18:25.919> however<00:18:26.160> we<00:18:26.320> want.<00:18:26.640> Can 00:18:26.789 --> 00:18:26.799 align:start position:0% to download and use however we want. Can 00:18:26.799 --> 00:18:28.310 align:start position:0% to download and use however we want. Can use<00:18:26.880> it<00:18:27.039> through<00:18:27.200> open<00:18:27.440> router<00:18:27.760> whenever<00:18:28.080> they 00:18:28.310 --> 00:18:28.320 align:start position:0% use it through open router whenever they 00:18:28.320 --> 00:18:31.510 align:start position:0% use it through open router whenever they put<00:18:28.799> 3x<00:18:29.440> additional<00:18:30.160> compute<00:18:30.640> into<00:18:30.960> the<00:18:31.200> post 00:18:31.510 --> 00:18:31.520 align:start position:0% put 3x additional compute into the post 00:18:31.520 --> 00:18:33.029 align:start position:0% put 3x additional compute into the post training<00:18:32.000> based<00:18:32.160> on<00:18:32.320> their<00:18:32.480> estimates.<00:18:32.799> I 00:18:33.029 --> 00:18:33.039 align:start position:0% training based on their estimates. I 00:18:33.039 --> 00:18:33.830 align:start position:0% training based on their estimates. I don't<00:18:33.120> know<00:18:33.120> if<00:18:33.280> they<00:18:33.440> actually<00:18:33.600> know<00:18:33.760> how 00:18:33.830 --> 00:18:33.840 align:start position:0% don't know if they actually know how 00:18:33.840 --> 00:18:35.029 align:start position:0% don't know if they actually know how much<00:18:34.000> compute<00:18:34.240> went<00:18:34.400> into<00:18:34.640> the<00:18:34.799> original 00:18:35.029 --> 00:18:35.039 align:start position:0% much compute went into the original 00:18:35.039 --> 00:18:36.230 align:start position:0% much compute went into the original model<00:18:35.280> or<00:18:35.440> not,<00:18:35.600> but<00:18:35.840> with<00:18:36.080> their 00:18:36.230 --> 00:18:36.240 align:start position:0% model or not, but with their 00:18:36.240 --> 00:18:39.510 align:start position:0% model or not, but with their assumptions,<00:18:37.039> they<00:18:37.200> did<00:18:37.440> 3x<00:18:38.080> more<00:18:38.400> on<00:18:38.640> top<00:18:39.280> in 00:18:39.510 --> 00:18:39.520 align:start position:0% assumptions, they did 3x more on top in 00:18:39.520 --> 00:18:41.510 align:start position:0% assumptions, they did 3x more on top in order<00:18:39.679> to<00:18:40.320> make<00:18:40.480> this<00:18:40.720> model<00:18:41.039> so<00:18:41.200> much<00:18:41.360> better 00:18:41.510 --> 00:18:41.520 align:start position:0% order to make this model so much better 00:18:41.520 --> 00:18:43.350 align:start position:0% order to make this model so much better at<00:18:41.760> code<00:18:42.000> and<00:18:42.240> also<00:18:42.559> worse<00:18:42.799> at<00:18:42.960> other<00:18:43.120> things. 00:18:43.350 --> 00:18:43.360 align:start position:0% at code and also worse at other things. 00:18:43.360 --> 00:18:44.470 align:start position:0% at code and also worse at other things. That's<00:18:43.520> why<00:18:43.679> their<00:18:43.760> evals<00:18:44.160> are<00:18:44.320> very 00:18:44.470 --> 00:18:44.480 align:start position:0% That's why their evals are very 00:18:44.480 --> 00:18:45.990 align:start position:0% That's why their evals are very different<00:18:44.640> cuz<00:18:44.799> the<00:18:44.960> model's<00:18:45.360> effectively<00:18:45.760> a 00:18:45.990 --> 00:18:46.000 align:start position:0% different cuz the model's effectively a 00:18:46.000 --> 00:18:47.350 align:start position:0% different cuz the model's effectively a whole<00:18:46.240> different<00:18:46.480> model.<00:18:46.880> I<00:18:47.039> really<00:18:47.200> want<00:18:47.280> to 00:18:47.350 --> 00:18:47.360 align:start position:0% whole different model. I really want to 00:18:47.360 --> 00:18:48.870 align:start position:0% whole different model. I really want to emphasize<00:18:47.679> how<00:18:47.840> big<00:18:47.919> of<00:18:48.000> a<00:18:48.160> deal<00:18:48.240> this<00:18:48.480> is. 00:18:48.870 --> 00:18:48.880 align:start position:0% emphasize how big of a deal this is. 00:18:48.880 --> 00:18:50.549 align:start position:0% emphasize how big of a deal this is. Most<00:18:49.120> of<00:18:49.280> the<00:18:49.440> major<00:18:49.760> model<00:18:50.080> releases<00:18:50.400> from 00:18:50.549 --> 00:18:50.559 align:start position:0% Most of the major model releases from 00:18:50.559 --> 00:18:52.470 align:start position:0% Most of the major model releases from the<00:18:50.720> major<00:18:50.960> labs<00:18:51.440> are<00:18:51.760> not<00:18:51.919> a<00:18:52.160> new 00:18:52.470 --> 00:18:52.480 align:start position:0% the major labs are not a new 00:18:52.480 --> 00:18:54.630 align:start position:0% the major labs are not a new pre-training<00:18:53.280> base.<00:18:53.840> The<00:18:54.080> pre-training<00:18:54.480> is 00:18:54.630 --> 00:18:54.640 align:start position:0% pre-training base. The pre-training is 00:18:54.640 --> 00:18:55.830 align:start position:0% pre-training base. The pre-training is the<00:18:54.799> first<00:18:54.880> step.<00:18:55.120> It's<00:18:55.360> almost<00:18:55.520> like<00:18:55.679> a 00:18:55.830 --> 00:18:55.840 align:start position:0% the first step. It's almost like a 00:18:55.840 --> 00:18:57.590 align:start position:0% the first step. It's almost like a template<00:18:56.320> in<00:18:56.559> a<00:18:56.720> way<00:18:56.799> and<00:18:57.039> then<00:18:57.120> they<00:18:57.360> do<00:18:57.440> a 00:18:57.590 --> 00:18:57.600 align:start position:0% template in a way and then they do a 00:18:57.600 --> 00:19:00.150 align:start position:0% template in a way and then they do a bunch<00:18:57.760> of<00:18:58.000> refinement<00:18:58.720> after<00:18:59.440> to<00:18:59.760> turn<00:18:59.919> that 00:19:00.150 --> 00:19:00.160 align:start position:0% bunch of refinement after to turn that 00:19:00.160 --> 00:19:02.310 align:start position:0% bunch of refinement after to turn that into<00:19:00.320> the<00:19:00.480> models<00:19:00.799> we<00:19:01.039> actually<00:19:01.360> like<00:19:01.600> to<00:19:01.840> use. 00:19:02.310 --> 00:19:02.320 align:start position:0% into the models we actually like to use. 00:19:02.320 --> 00:19:03.830 align:start position:0% into the models we actually like to use. Most<00:19:02.640> models<00:19:02.960> start<00:19:03.200> as<00:19:03.440> like<00:19:03.600> an 00:19:03.830 --> 00:19:03.840 align:start position:0% Most models start as like an 00:19:03.840 --> 00:19:05.110 align:start position:0% Most models start as like an autocomplete<00:19:04.400> model<00:19:04.559> with<00:19:04.720> a<00:19:04.880> bunch<00:19:04.960> of 00:19:05.110 --> 00:19:05.120 align:start position:0% autocomplete model with a bunch of 00:19:05.120 --> 00:19:06.390 align:start position:0% autocomplete model with a bunch of information<00:19:05.360> in<00:19:05.600> them,<00:19:05.840> but<00:19:06.000> they're<00:19:06.240> not 00:19:06.390 --> 00:19:06.400 align:start position:0% information in them, but they're not 00:19:06.400 --> 00:19:07.909 align:start position:0% information in them, but they're not good<00:19:06.480> at<00:19:06.640> talking<00:19:06.960> or<00:19:07.120> tool<00:19:07.440> calling<00:19:07.679> or<00:19:07.840> all 00:19:07.909 --> 00:19:07.919 align:start position:0% good at talking or tool calling or all 00:19:07.919 --> 00:19:08.950 align:start position:0% good at talking or tool calling or all these<00:19:08.160> other<00:19:08.240> things,<00:19:08.640> and<00:19:08.799> then 00:19:08.950 --> 00:19:08.960 align:start position:0% these other things, and then 00:19:08.960 --> 00:19:10.630 align:start position:0% these other things, and then post-training<00:19:09.679> brings<00:19:09.919> those<00:19:10.160> behaviors 00:19:10.630 --> 00:19:10.640 align:start position:0% post-training brings those behaviors 00:19:10.640 --> 00:19:13.190 align:start position:0% post-training brings those behaviors out.<00:19:11.039> Models<00:19:11.360> like<00:19:11.440> GBT<00:19:12.000> 5<00:19:12.080> and<00:19:12.320> 5.1<00:19:12.960> were 00:19:13.190 --> 00:19:13.200 align:start position:0% out. Models like GBT 5 and 5.1 were 00:19:13.200 --> 00:19:15.190 align:start position:0% out. Models like GBT 5 and 5.1 were based<00:19:13.440> on<00:19:13.520> the<00:19:13.679> same<00:19:13.840> pre-training.<00:19:14.559> 52 00:19:15.190 --> 00:19:15.200 align:start position:0% based on the same pre-training. 52 00:19:15.200 --> 00:19:17.110 align:start position:0% based on the same pre-training. 52 onwards<00:19:15.760> also<00:19:16.080> seem<00:19:16.240> to<00:19:16.400> all<00:19:16.559> be<00:19:16.720> the<00:19:16.960> same 00:19:17.110 --> 00:19:17.120 align:start position:0% onwards also seem to all be the same 00:19:17.120 --> 00:19:19.430 align:start position:0% onwards also seem to all be the same pre-training.<00:19:17.760> Gemini<00:19:18.160> 3<00:19:18.400> to<00:19:18.559> 3.1<00:19:19.120> was<00:19:19.280> a 00:19:19.430 --> 00:19:19.440 align:start position:0% pre-training. Gemini 3 to 3.1 was a 00:19:19.440 --> 00:19:20.870 align:start position:0% pre-training. Gemini 3 to 3.1 was a post-training<00:19:19.919> change.<00:19:20.240> The<00:19:20.400> pre-training 00:19:20.870 --> 00:19:20.880 align:start position:0% post-training change. The pre-training 00:19:20.880 --> 00:19:22.870 align:start position:0% post-training change. The pre-training was<00:19:20.960> the<00:19:21.120> same.<00:19:21.360> It<00:19:21.520> is<00:19:21.679> very<00:19:22.000> common<00:19:22.320> for<00:19:22.559> labs 00:19:22.870 --> 00:19:22.880 align:start position:0% was the same. It is very common for labs 00:19:22.880 --> 00:19:25.110 align:start position:0% was the same. It is very common for labs to<00:19:23.039> take<00:19:23.280> existing<00:19:23.840> models<00:19:24.400> and<00:19:24.720> use 00:19:25.110 --> 00:19:25.120 align:start position:0% to take existing models and use 00:19:25.120 --> 00:19:26.789 align:start position:0% to take existing models and use post-training<00:19:25.760> to<00:19:26.000> make<00:19:26.160> them<00:19:26.320> behave 00:19:26.789 --> 00:19:26.799 align:start position:0% post-training to make them behave 00:19:26.799 --> 00:19:28.789 align:start position:0% post-training to make them behave entirely<00:19:27.440> differently<00:19:27.840> in<00:19:28.080> ways<00:19:28.320> that<00:19:28.480> are 00:19:28.789 --> 00:19:28.799 align:start position:0% entirely differently in ways that are 00:19:28.799 --> 00:19:31.270 align:start position:0% entirely differently in ways that are often<00:19:29.520> massive<00:19:29.919> wins.<00:19:30.400> What<00:19:30.640> we<00:19:30.799> haven't<00:19:31.039> seen 00:19:31.270 --> 00:19:31.280 align:start position:0% often massive wins. What we haven't seen 00:19:31.280 --> 00:19:33.110 align:start position:0% often massive wins. What we haven't seen before<00:19:31.600> though<00:19:31.840> is<00:19:32.000> a<00:19:32.160> lab<00:19:32.400> using<00:19:32.720> post 00:19:33.110 --> 00:19:33.120 align:start position:0% before though is a lab using post 00:19:33.120 --> 00:19:34.950 align:start position:0% before though is a lab using post training<00:19:33.520> to<00:19:33.760> put<00:19:33.919> out<00:19:34.160> something<00:19:34.400> that<00:19:34.640> gets 00:19:34.950 --> 00:19:34.960 align:start position:0% training to put out something that gets 00:19:34.960 --> 00:19:37.029 align:start position:0% training to put out something that gets frontier<00:19:35.440> level<00:19:35.679> numbers<00:19:36.400> by<00:19:36.640> doing<00:19:36.799> this 00:19:37.029 --> 00:19:37.039 align:start position:0% frontier level numbers by doing this 00:19:37.039 --> 00:19:39.190 align:start position:0% frontier level numbers by doing this post-<00:19:37.360> training<00:19:37.600> on<00:19:37.919> someone<00:19:38.240> else's<00:19:38.720> model. 00:19:39.190 --> 00:19:39.200 align:start position:0% post- training on someone else's model. 00:19:39.200 --> 00:19:40.870 align:start position:0% post- training on someone else's model. But<00:19:39.360> to<00:19:39.520> go<00:19:39.679> back<00:19:39.840> to<00:19:40.000> a<00:19:40.240> thing<00:19:40.320> I<00:19:40.559> mentioned 00:19:40.870 --> 00:19:40.880 align:start position:0% But to go back to a thing I mentioned 00:19:40.880 --> 00:19:43.270 align:start position:0% But to go back to a thing I mentioned earlier<00:19:41.520> about<00:19:42.000> this<00:19:42.400> license<00:19:42.880> and<00:19:43.120> in 00:19:43.270 --> 00:19:43.280 align:start position:0% earlier about this license and in 00:19:43.280 --> 00:19:45.830 align:start position:0% earlier about this license and in particular<00:19:43.840> the<00:19:44.160> ability<00:19:44.480> to<00:19:44.799> obuscate<00:19:45.520> who's 00:19:45.830 --> 00:19:45.840 align:start position:0% particular the ability to obuscate who's 00:19:45.840 --> 00:19:48.070 align:start position:0% particular the ability to obuscate who's responsible<00:19:46.400> by<00:19:46.640> having<00:19:46.799> a<00:19:46.960> middleman.<00:19:47.520> Lero 00:19:48.070 --> 00:19:48.080 align:start position:0% responsible by having a middleman. Lero 00:19:48.080 --> 00:19:50.070 align:start position:0% responsible by having a middleman. Lero says<00:19:48.240> here<00:19:48.640> we<00:19:48.799> are<00:19:48.960> following<00:19:49.280> the<00:19:49.440> license 00:19:50.070 --> 00:19:50.080 align:start position:0% says here we are following the license 00:19:50.080 --> 00:19:52.950 align:start position:0% says here we are following the license through<00:19:50.480> our<00:19:50.880> inference<00:19:51.520> partner<00:19:52.240> terms. 00:19:52.950 --> 00:19:52.960 align:start position:0% through our inference partner terms. 00:19:52.960 --> 00:19:54.310 align:start position:0% through our inference partner terms. Sometimes<00:19:53.360> being<00:19:53.520> asserted<00:19:53.840> into<00:19:54.000> the<00:19:54.160> middle 00:19:54.310 --> 00:19:54.320 align:start position:0% Sometimes being asserted into the middle 00:19:54.320 --> 00:19:55.669 align:start position:0% Sometimes being asserted into the middle of<00:19:54.480> something<00:19:54.720> like<00:19:54.880> this<00:19:55.039> is<00:19:55.280> good.<00:19:55.520> Just 00:19:55.669 --> 00:19:55.679 align:start position:0% of something like this is good. Just 00:19:55.679 --> 00:19:56.950 align:start position:0% of something like this is good. Just like<00:19:55.760> we're<00:19:55.919> about<00:19:56.080> to<00:19:56.160> insert<00:19:56.400> an<00:19:56.559> ad<00:19:56.720> in<00:19:56.880> the 00:19:56.950 --> 00:19:56.960 align:start position:0% like we're about to insert an ad in the 00:19:56.960 --> 00:19:58.390 align:start position:0% like we're about to insert an ad in the middle<00:19:57.039> of<00:19:57.120> this<00:19:57.280> video.<00:19:57.520> Engineers<00:19:58.080> hate 00:19:58.390 --> 00:19:58.400 align:start position:0% middle of this video. Engineers hate 00:19:58.400 --> 00:19:59.909 align:start position:0% middle of this video. Engineers hate meetings,<00:19:58.799> but<00:19:58.960> there's<00:19:59.200> one<00:19:59.360> thing<00:19:59.520> I<00:19:59.760> know 00:19:59.909 --> 00:19:59.919 align:start position:0% meetings, but there's one thing I know 00:19:59.919 --> 00:20:02.310 align:start position:0% meetings, but there's one thing I know they<00:20:00.160> hate<00:20:00.400> way,<00:20:00.720> way<00:20:01.039> more.<00:20:01.600> Interviews. 00:20:02.310 --> 00:20:02.320 align:start position:0% they hate way, way more. Interviews. 00:20:02.320 --> 00:20:04.070 align:start position:0% they hate way, way more. Interviews. Especially<00:20:02.720> with<00:20:02.960> bad<00:20:03.200> candidates.<00:20:03.760> And<00:20:03.919> I'll 00:20:04.070 --> 00:20:04.080 align:start position:0% Especially with bad candidates. And I'll 00:20:04.080 --> 00:20:05.590 align:start position:0% Especially with bad candidates. And I'll be<00:20:04.160> real,<00:20:04.480> it's<00:20:04.640> hard<00:20:04.799> to<00:20:04.960> find<00:20:05.120> the<00:20:05.280> good<00:20:05.440> ones 00:20:05.590 --> 00:20:05.600 align:start position:0% be real, it's hard to find the good ones 00:20:05.600 --> 00:20:07.510 align:start position:0% be real, it's hard to find the good ones now.<00:20:05.840> With<00:20:06.000> all<00:20:06.080> the<00:20:06.240> AI<00:20:06.559> slop<00:20:06.960> clogging<00:20:07.360> up 00:20:07.510 --> 00:20:07.520 align:start position:0% now. With all the AI slop clogging up 00:20:07.520 --> 00:20:09.430 align:start position:0% now. With all the AI slop clogging up our<00:20:07.679> inbound<00:20:08.080> forms,<00:20:08.480> it's<00:20:08.720> just<00:20:08.960> impossible 00:20:09.430 --> 00:20:09.440 align:start position:0% our inbound forms, it's just impossible 00:20:09.440 --> 00:20:11.029 align:start position:0% our inbound forms, it's just impossible to<00:20:09.600> get<00:20:09.760> the<00:20:09.919> good<00:20:10.160> people<00:20:10.400> out.<00:20:10.799> Unless 00:20:11.029 --> 00:20:11.039 align:start position:0% to get the good people out. Unless 00:20:11.039 --> 00:20:13.270 align:start position:0% to get the good people out. Unless you're<00:20:11.200> using<00:20:11.440> today's<00:20:11.760> sponsor,<00:20:12.160> G2I.<00:20:12.960> These 00:20:13.270 --> 00:20:13.280 align:start position:0% you're using today's sponsor, G2I. These 00:20:13.280 --> 00:20:14.950 align:start position:0% you're using today's sponsor, G2I. These guys<00:20:13.600> get<00:20:13.919> hiring.<00:20:14.400> They<00:20:14.480> have<00:20:14.559> an<00:20:14.720> incredible 00:20:14.950 --> 00:20:14.960 align:start position:0% guys get hiring. They have an incredible 00:20:14.960 --> 00:20:16.470 align:start position:0% guys get hiring. They have an incredible network<00:20:15.200> of<00:20:15.360> engineers<00:20:15.760> ready<00:20:15.919> to<00:20:16.000> go<00:20:16.160> and<00:20:16.320> one 00:20:16.470 --> 00:20:16.480 align:start position:0% network of engineers ready to go and one 00:20:16.480 --> 00:20:18.150 align:start position:0% network of engineers ready to go and one of<00:20:16.559> the<00:20:16.720> best<00:20:16.880> possible<00:20:17.360> platforms<00:20:17.840> to<00:20:18.000> find 00:20:18.150 --> 00:20:18.160 align:start position:0% of the best possible platforms to find 00:20:18.160 --> 00:20:19.669 align:start position:0% of the best possible platforms to find who's<00:20:18.400> the<00:20:18.559> right<00:20:18.720> fit.<00:20:19.120> You<00:20:19.280> talk<00:20:19.360> with<00:20:19.520> them 00:20:19.669 --> 00:20:19.679 align:start position:0% who's the right fit. You talk with them 00:20:19.679 --> 00:20:20.789 align:start position:0% who's the right fit. You talk with them about<00:20:19.760> what<00:20:19.919> you're<00:20:20.080> looking<00:20:20.240> for.<00:20:20.559> You<00:20:20.640> send 00:20:20.789 --> 00:20:20.799 align:start position:0% about what you're looking for. You send 00:20:20.799 --> 00:20:21.909 align:start position:0% about what you're looking for. You send them<00:20:20.960> some<00:20:21.120> questions<00:20:21.360> to<00:20:21.600> ask<00:20:21.760> the 00:20:21.909 --> 00:20:21.919 align:start position:0% them some questions to ask the 00:20:21.919 --> 00:20:23.750 align:start position:0% them some questions to ask the candidates<00:20:22.320> and<00:20:22.480> then<00:20:22.640> they<00:20:22.799> send<00:20:22.960> you<00:20:23.360> videos 00:20:23.750 --> 00:20:23.760 align:start position:0% candidates and then they send you videos 00:20:23.760 --> 00:20:25.750 align:start position:0% candidates and then they send you videos of<00:20:24.000> the<00:20:24.240> candidates<00:20:24.640> responding<00:20:25.120> to<00:20:25.360> try<00:20:25.600> and 00:20:25.750 --> 00:20:25.760 align:start position:0% of the candidates responding to try and 00:20:25.760 --> 00:20:27.430 align:start position:0% of the candidates responding to try and find<00:20:25.919> the<00:20:26.160> right<00:20:26.320> vibes.<00:20:26.720> I<00:20:26.960> know<00:20:27.039> a<00:20:27.200> ton<00:20:27.280> of 00:20:27.430 --> 00:20:27.440 align:start position:0% find the right vibes. I know a ton of 00:20:27.440 --> 00:20:28.870 align:start position:0% find the right vibes. I know a ton of candidates<00:20:27.760> and<00:20:28.000> companies<00:20:28.320> that<00:20:28.480> have<00:20:28.640> had<00:20:28.720> a 00:20:28.870 --> 00:20:28.880 align:start position:0% candidates and companies that have had a 00:20:28.880 --> 00:20:30.470 align:start position:0% candidates and companies that have had a great<00:20:29.039> time<00:20:29.120> with<00:20:29.280> G2I<00:20:29.840> and<00:20:30.000> that's<00:20:30.159> why<00:20:30.240> big 00:20:30.470 --> 00:20:30.480 align:start position:0% great time with G2I and that's why big 00:20:30.480 --> 00:20:32.070 align:start position:0% great time with G2I and that's why big companies<00:20:30.720> like<00:20:30.960> Web<00:20:31.120> Flow,<00:20:31.440> One<00:20:31.679> Password 00:20:32.070 --> 00:20:32.080 align:start position:0% companies like Web Flow, One Password 00:20:32.080 --> 00:20:33.990 align:start position:0% companies like Web Flow, One Password Automatic,<00:20:32.559> and<00:20:32.799> Meta<00:20:33.280> are<00:20:33.440> all<00:20:33.600> working<00:20:33.840> with 00:20:33.990 --> 00:20:34.000 align:start position:0% Automatic, and Meta are all working with 00:20:34.000 --> 00:20:35.669 align:start position:0% Automatic, and Meta are all working with them<00:20:34.159> to<00:20:34.400> hire<00:20:34.640> the<00:20:34.880> best<00:20:35.039> engineers.<00:20:35.520> I 00:20:35.669 --> 00:20:35.679 align:start position:0% them to hire the best engineers. I 00:20:35.679 --> 00:20:36.870 align:start position:0% them to hire the best engineers. I wouldn't<00:20:35.840> believe<00:20:36.080> this<00:20:36.240> either,<00:20:36.480> but<00:20:36.640> I've 00:20:36.870 --> 00:20:36.880 align:start position:0% wouldn't believe this either, but I've 00:20:36.880 --> 00:20:38.710 align:start position:0% wouldn't believe this either, but I've hung<00:20:37.039> out<00:20:37.120> with<00:20:37.280> these<00:20:37.440> guys<00:20:37.679> a<00:20:37.919> ton.<00:20:38.320> They<00:20:38.559> do 00:20:38.710 --> 00:20:38.720 align:start position:0% hung out with these guys a ton. They do 00:20:38.720 --> 00:20:40.549 align:start position:0% hung out with these guys a ton. They do really<00:20:38.960> get<00:20:39.120> it.<00:20:39.360> React<00:20:39.760> Miami<00:20:40.080> is<00:20:40.320> my 00:20:40.549 --> 00:20:40.559 align:start position:0% really get it. React Miami is my 00:20:40.559 --> 00:20:42.630 align:start position:0% really get it. React Miami is my favorite<00:20:40.799> developer<00:20:41.280> conference<00:20:41.760> by<00:20:42.159> far, 00:20:42.630 --> 00:20:42.640 align:start position:0% favorite developer conference by far, 00:20:42.640 --> 00:20:44.070 align:start position:0% favorite developer conference by far, but<00:20:42.799> they<00:20:42.960> might<00:20:43.200> want<00:20:43.360> up<00:20:43.520> it<00:20:43.679> this<00:20:43.840> year 00:20:44.070 --> 00:20:44.080 align:start position:0% but they might want up it this year 00:20:44.080 --> 00:20:45.669 align:start position:0% but they might want up it this year because<00:20:44.320> they<00:20:44.559> just<00:20:44.720> announced<00:20:45.039> the<00:20:45.280> AI 00:20:45.669 --> 00:20:45.679 align:start position:0% because they just announced the AI 00:20:45.679 --> 00:20:48.070 align:start position:0% because they just announced the AI engineer<00:20:46.320> Miami<00:20:46.880> conference.<00:20:47.520> Yes,<00:20:47.840> they're 00:20:48.070 --> 00:20:48.080 align:start position:0% engineer Miami conference. Yes, they're 00:20:48.080 --> 00:20:49.909 align:start position:0% engineer Miami conference. Yes, they're bringing<00:20:48.400> AI<00:20:48.720> to<00:20:48.880> Miami.<00:20:49.440> Sure,<00:20:49.600> a<00:20:49.760> lot<00:20:49.840> of 00:20:49.909 --> 00:20:49.919 align:start position:0% bringing AI to Miami. Sure, a lot of 00:20:49.919 --> 00:20:51.430 align:start position:0% bringing AI to Miami. Sure, a lot of people<00:20:50.080> will<00:20:50.240> hate<00:20:50.400> it,<00:20:50.559> but<00:20:50.799> us<00:20:50.960> developers 00:20:51.430 --> 00:20:51.440 align:start position:0% people will hate it, but us developers 00:20:51.440 --> 00:20:52.789 align:start position:0% people will hate it, but us developers are<00:20:51.679> going<00:20:51.760> to<00:20:51.840> love<00:20:52.000> the<00:20:52.240> excuse<00:20:52.480> to<00:20:52.640> be 00:20:52.789 --> 00:20:52.799 align:start position:0% are going to love the excuse to be 00:20:52.799 --> 00:20:54.390 align:start position:0% are going to love the excuse to be there.<00:20:53.039> So,<00:20:53.200> if<00:20:53.360> you're<00:20:53.520> in<00:20:53.679> Florida<00:20:54.000> or<00:20:54.240> you 00:20:54.390 --> 00:20:54.400 align:start position:0% there. So, if you're in Florida or you 00:20:54.400 --> 00:20:56.390 align:start position:0% there. So, if you're in Florida or you can<00:20:54.559> be<00:20:54.720> for<00:20:54.960> April<00:20:55.200> 20th<00:20:55.440> and<00:20:55.679> 21st,<00:20:56.159> I<00:20:56.320> would 00:20:56.390 --> 00:20:56.400 align:start position:0% can be for April 20th and 21st, I would 00:20:56.400 --> 00:20:57.750 align:start position:0% can be for April 20th and 21st, I would highly<00:20:56.640> recommend<00:20:56.960> being<00:20:57.200> there.<00:20:57.520> You<00:20:57.679> can 00:20:57.750 --> 00:20:57.760 align:start position:0% highly recommend being there. You can 00:20:57.760 --> 00:20:58.710 align:start position:0% highly recommend being there. You can bet<00:20:57.919> your<00:20:58.000> ass<00:20:58.240> you're<00:20:58.400> going<00:20:58.400> to<00:20:58.480> see<00:20:58.640> me 00:20:58.710 --> 00:20:58.720 align:start position:0% bet your ass you're going to see me 00:20:58.720 --> 00:21:00.070 align:start position:0% bet your ass you're going to see me there,<00:20:58.960> too.<00:20:59.200> Hire<00:20:59.440> some<00:20:59.600> great<00:20:59.760> engineers 00:21:00.070 --> 00:21:00.080 align:start position:0% there, too. Hire some great engineers 00:21:00.080 --> 00:21:02.789 align:start position:0% there, too. Hire some great engineers and<00:21:00.240> maybe<00:21:00.559> come<00:21:00.720> party<00:21:00.960> at<00:21:01.120> soyv.link/g2i. 00:21:02.789 --> 00:21:02.799 align:start position:0% and maybe come party at soyv.link/g2i. 00:21:02.799 --> 00:21:04.549 align:start position:0% and maybe come party at soyv.link/g2i. Sorry<00:21:02.960> about<00:21:03.120> that.<00:21:03.360> As<00:21:03.600> I<00:21:03.760> was<00:21:03.919> saying,<00:21:04.400> the 00:21:04.549 --> 00:21:04.559 align:start position:0% Sorry about that. As I was saying, the 00:21:04.559 --> 00:21:06.630 align:start position:0% Sorry about that. As I was saying, the model<00:21:05.039> has<00:21:05.200> an<00:21:05.520> inference<00:21:05.919> partner<00:21:06.400> that 00:21:06.630 --> 00:21:06.640 align:start position:0% model has an inference partner that 00:21:06.640 --> 00:21:08.870 align:start position:0% model has an inference partner that allowed<00:21:07.039> cursor<00:21:07.600> to<00:21:07.760> not<00:21:08.000> have<00:21:08.159> to<00:21:08.320> disclose 00:21:08.870 --> 00:21:08.880 align:start position:0% allowed cursor to not have to disclose 00:21:08.880 --> 00:21:11.750 align:start position:0% allowed cursor to not have to disclose by<00:21:09.200> saying<00:21:09.440> Kimmy<00:21:09.840> K2.5<00:21:10.559> in<00:21:10.720> the<00:21:10.880> UI.<00:21:11.280> Lero 00:21:11.750 --> 00:21:11.760 align:start position:0% by saying Kimmy K2.5 in the UI. Lero 00:21:11.760 --> 00:21:13.430 align:start position:0% by saying Kimmy K2.5 in the UI. Lero replied<00:21:12.080> here<00:21:12.320> since<00:21:12.640> people<00:21:12.880> really<00:21:13.120> want<00:21:13.360> me 00:21:13.430 --> 00:21:13.440 align:start position:0% replied here since people really want me 00:21:13.440 --> 00:21:16.470 align:start position:0% replied here since people really want me to<00:21:13.600> say<00:21:13.760> this.<00:21:14.080> Kimmy<00:21:14.559> K2.5.<00:21:15.760> Yes,<00:21:16.159> that<00:21:16.400> is 00:21:16.470 --> 00:21:16.480 align:start position:0% to say this. Kimmy K2.5. Yes, that is 00:21:16.480 --> 00:21:18.149 align:start position:0% to say this. Kimmy K2.5. Yes, that is the<00:21:16.640> base<00:21:16.880> we<00:21:17.120> started<00:21:17.360> from<00:21:17.600> and<00:21:17.840> we<00:21:18.000> are 00:21:18.149 --> 00:21:18.159 align:start position:0% the base we started from and we are 00:21:18.159 --> 00:21:19.830 align:start position:0% the base we started from and we are following<00:21:18.400> the<00:21:18.559> license<00:21:19.039> through<00:21:19.360> inference 00:21:19.830 --> 00:21:19.840 align:start position:0% following the license through inference 00:21:19.840 --> 00:21:21.990 align:start position:0% following the license through inference partner<00:21:20.159> terms.<00:21:20.640> In<00:21:20.799> this<00:21:20.960> case,<00:21:21.280> fireworks. 00:21:21.990 --> 00:21:22.000 align:start position:0% partner terms. In this case, fireworks. 00:21:22.000 --> 00:21:23.350 align:start position:0% partner terms. In this case, fireworks. If<00:21:22.159> you're<00:21:22.240> not<00:21:22.400> familiar<00:21:22.559> with<00:21:22.720> Fireworks, 00:21:23.350 --> 00:21:23.360 align:start position:0% If you're not familiar with Fireworks, 00:21:23.360 --> 00:21:25.510 align:start position:0% If you're not familiar with Fireworks, they're<00:21:23.600> a<00:21:23.840> compute<00:21:24.240> platform<00:21:24.880> for<00:21:25.120> hosting 00:21:25.510 --> 00:21:25.520 align:start position:0% they're a compute platform for hosting 00:21:25.520 --> 00:21:27.750 align:start position:0% they're a compute platform for hosting openweight<00:21:26.159> models<00:21:26.559> and<00:21:26.880> using<00:21:27.120> them<00:21:27.360> to<00:21:27.520> do 00:21:27.750 --> 00:21:27.760 align:start position:0% openweight models and using them to do 00:21:27.760 --> 00:21:29.510 align:start position:0% openweight models and using them to do fast<00:21:28.080> inference.<00:21:28.559> They<00:21:28.720> also<00:21:28.880> provide<00:21:29.120> info 00:21:29.510 --> 00:21:29.520 align:start position:0% fast inference. They also provide info 00:21:29.520 --> 00:21:30.950 align:start position:0% fast inference. They also provide info for<00:21:29.600> people<00:21:29.760> who<00:21:29.919> want<00:21:30.000> to<00:21:30.080> do<00:21:30.240> their<00:21:30.400> own<00:21:30.559> post 00:21:30.950 --> 00:21:30.960 align:start position:0% for people who want to do their own post 00:21:30.960 --> 00:21:32.470 align:start position:0% for people who want to do their own post training,<00:21:31.200> which<00:21:31.520> seems<00:21:31.760> to<00:21:31.840> be<00:21:32.000> where<00:21:32.240> a<00:21:32.400> lot 00:21:32.470 --> 00:21:32.480 align:start position:0% training, which seems to be where a lot 00:21:32.480 --> 00:21:34.470 align:start position:0% training, which seems to be where a lot of<00:21:32.559> the<00:21:32.720> inference<00:21:33.120> that<00:21:33.600> cursor<00:21:34.000> was<00:21:34.159> running 00:21:34.470 --> 00:21:34.480 align:start position:0% of the inference that cursor was running 00:21:34.480 --> 00:21:36.549 align:start position:0% of the inference that cursor was running was<00:21:34.799> run.<00:21:35.280> And<00:21:35.520> because<00:21:35.840> people<00:21:36.159> were<00:21:36.320> still 00:21:36.549 --> 00:21:36.559 align:start position:0% was run. And because people were still 00:21:36.559 --> 00:21:39.110 align:start position:0% was run. And because people were still speculating<00:21:37.200> about<00:21:37.520> this,<00:21:38.159> eventually<00:21:38.640> Kimmy 00:21:39.110 --> 00:21:39.120 align:start position:0% speculating about this, eventually Kimmy 00:21:39.120 --> 00:21:41.270 align:start position:0% speculating about this, eventually Kimmy came<00:21:39.280> out<00:21:39.520> and<00:21:39.919> said<00:21:40.320> what<00:21:40.720> effectively 00:21:41.270 --> 00:21:41.280 align:start position:0% came out and said what effectively 00:21:41.280 --> 00:21:43.110 align:start position:0% came out and said what effectively Cursor<00:21:41.760> was<00:21:41.919> saying<00:21:42.080> as<00:21:42.240> well.<00:21:42.559> Congrats<00:21:42.960> to 00:21:43.110 --> 00:21:43.120 align:start position:0% Cursor was saying as well. Congrats to 00:21:43.120 --> 00:21:45.029 align:start position:0% Cursor was saying as well. Congrats to the<00:21:43.200> Cursor<00:21:43.600> AI<00:21:44.000> team<00:21:44.159> on<00:21:44.400> the<00:21:44.559> launch<00:21:44.799> of 00:21:45.029 --> 00:21:45.039 align:start position:0% the Cursor AI team on the launch of 00:21:45.039 --> 00:21:47.430 align:start position:0% the Cursor AI team on the launch of Composer<00:21:45.679> 2.<00:21:46.080> We<00:21:46.320> are<00:21:46.480> proud<00:21:46.640> to<00:21:46.799> see<00:21:47.039> Kimmy 00:21:47.430 --> 00:21:47.440 align:start position:0% Composer 2. We are proud to see Kimmy 00:21:47.440 --> 00:21:49.669 align:start position:0% Composer 2. We are proud to see Kimmy K25<00:21:48.159> provide<00:21:48.480> the<00:21:48.720> foundation.<00:21:49.280> Seeing<00:21:49.520> our 00:21:49.669 --> 00:21:49.679 align:start position:0% K25 provide the foundation. Seeing our 00:21:49.679 --> 00:21:50.789 align:start position:0% K25 provide the foundation. Seeing our model<00:21:49.840> integrated<00:21:50.240> effectively<00:21:50.640> through 00:21:50.789 --> 00:21:50.799 align:start position:0% model integrated effectively through 00:21:50.799 --> 00:21:52.549 align:start position:0% model integrated effectively through Cursor's<00:21:51.360> continued<00:21:51.679> pre-training<00:21:52.320> and 00:21:52.549 --> 00:21:52.559 align:start position:0% Cursor's continued pre-training and 00:21:52.559 --> 00:21:54.870 align:start position:0% Cursor's continued pre-training and highMP<00:21:53.200> RL<00:21:53.520> training<00:21:53.919> is<00:21:54.159> the<00:21:54.320> open<00:21:54.640> model 00:21:54.870 --> 00:21:54.880 align:start position:0% highMP RL training is the open model 00:21:54.880 --> 00:21:57.110 align:start position:0% highMP RL training is the open model ecosystem<00:21:55.440> we<00:21:55.679> love<00:21:55.840> to<00:21:56.080> support.<00:21:56.720> Note 00:21:57.110 --> 00:21:57.120 align:start position:0% ecosystem we love to support. Note 00:21:57.120 --> 00:21:59.430 align:start position:0% ecosystem we love to support. Note Cursor<00:21:57.600> accessed<00:21:58.000> Kimkate<00:21:58.559> 2.5<00:21:59.120> via 00:21:59.430 --> 00:21:59.440 align:start position:0% Cursor accessed Kimkate 2.5 via 00:21:59.440 --> 00:22:02.230 align:start position:0% Cursor accessed Kimkate 2.5 via Fireworks<00:22:00.000> AI's<00:22:00.720> hosted<00:22:01.280> RL<00:22:01.679> and<00:22:01.840> inference 00:22:02.230 --> 00:22:02.240 align:start position:0% Fireworks AI's hosted RL and inference 00:22:02.240 --> 00:22:04.470 align:start position:0% Fireworks AI's hosted RL and inference platforms<00:22:02.960> as<00:22:03.280> part<00:22:03.440> of<00:22:03.520> an<00:22:03.760> authorized 00:22:04.470 --> 00:22:04.480 align:start position:0% platforms as part of an authorized 00:22:04.480 --> 00:22:06.630 align:start position:0% platforms as part of an authorized commercial<00:22:05.039> product<00:22:05.520> in<00:22:05.840> partnership.<00:22:06.480> This 00:22:06.630 --> 00:22:06.640 align:start position:0% commercial product in partnership. This 00:22:06.640 --> 00:22:09.029 align:start position:0% commercial product in partnership. This is<00:22:06.720> the<00:22:06.960> important<00:22:07.280> piece.<00:22:07.840> Since<00:22:08.320> Cursor<00:22:08.799> is 00:22:09.029 --> 00:22:09.039 align:start position:0% is the important piece. Since Cursor is 00:22:09.039 --> 00:22:11.590 align:start position:0% is the important piece. Since Cursor is going<00:22:09.280> through<00:22:09.679> Fireworks<00:22:10.480> and<00:22:10.799> Fireworks<00:22:11.360> is 00:22:11.590 --> 00:22:11.600 align:start position:0% going through Fireworks and Fireworks is 00:22:11.600 --> 00:22:13.750 align:start position:0% going through Fireworks and Fireworks is disclosing<00:22:12.159> properly<00:22:12.880> to<00:22:13.120> Cursor<00:22:13.600> that 00:22:13.750 --> 00:22:13.760 align:start position:0% disclosing properly to Cursor that 00:22:13.760 --> 00:22:16.470 align:start position:0% disclosing properly to Cursor that they're<00:22:13.919> using<00:22:14.080> Kimk<00:22:14.640> 2.5.<00:22:15.520> this<00:22:15.840> is<00:22:16.080> how<00:22:16.240> they 00:22:16.470 --> 00:22:16.480 align:start position:0% they're using Kimk 2.5. this is how they 00:22:16.480 --> 00:22:18.549 align:start position:0% they're using Kimk 2.5. this is how they get<00:22:16.640> around<00:22:17.039> that<00:22:17.280> license.<00:22:18.000> I<00:22:18.159> knew<00:22:18.320> this 00:22:18.549 --> 00:22:18.559 align:start position:0% get around that license. I knew this 00:22:18.559 --> 00:22:20.390 align:start position:0% get around that license. I knew this would<00:22:18.720> be<00:22:18.880> a<00:22:19.039> thing.<00:22:19.360> In<00:22:19.440> fact,<00:22:19.600> I<00:22:19.840> speculated 00:22:20.390 --> 00:22:20.400 align:start position:0% would be a thing. In fact, I speculated 00:22:20.400 --> 00:22:22.390 align:start position:0% would be a thing. In fact, I speculated as<00:22:20.559> much<00:22:20.799> in<00:22:21.120> the<00:22:21.280> past<00:22:21.600> that<00:22:21.840> this<00:22:22.080> type<00:22:22.240> of 00:22:22.390 --> 00:22:22.400 align:start position:0% as much in the past that this type of 00:22:22.400 --> 00:22:24.310 align:start position:0% as much in the past that this type of thing<00:22:22.640> would<00:22:22.880> allow<00:22:23.120> for<00:22:23.280> people<00:22:23.520> to<00:22:23.919> hide 00:22:24.310 --> 00:22:24.320 align:start position:0% thing would allow for people to hide 00:22:24.320 --> 00:22:25.990 align:start position:0% thing would allow for people to hide what<00:22:24.559> model<00:22:24.880> they're<00:22:25.120> using,<00:22:25.520> even<00:22:25.679> if<00:22:25.840> it's 00:22:25.990 --> 00:22:26.000 align:start position:0% what model they're using, even if it's 00:22:26.000 --> 00:22:27.990 align:start position:0% what model they're using, even if it's not<00:22:26.159> within<00:22:26.400> the<00:22:26.559> spirit<00:22:26.799> of<00:22:26.960> the<00:22:27.120> license.<00:22:27.760> I 00:22:27.990 --> 00:22:28.000 align:start position:0% not within the spirit of the license. I 00:22:28.000 --> 00:22:30.149 align:start position:0% not within the spirit of the license. I think<00:22:28.080> it's<00:22:28.400> important<00:22:28.720> for<00:22:29.039> businesses<00:22:29.760> to 00:22:30.149 --> 00:22:30.159 align:start position:0% think it's important for businesses to 00:22:30.159 --> 00:22:32.710 align:start position:0% think it's important for businesses to honor<00:22:30.559> the<00:22:30.880> spirit<00:22:31.200> of<00:22:31.360> these<00:22:31.679> licenses<00:22:32.400> in 00:22:32.710 --> 00:22:32.720 align:start position:0% honor the spirit of these licenses in 00:22:32.720 --> 00:22:35.430 align:start position:0% honor the spirit of these licenses in general.<00:22:33.360> We're<00:22:33.679> still<00:22:33.919> so<00:22:34.320> early<00:22:34.799> in<00:22:35.120> all<00:22:35.360> of 00:22:35.430 --> 00:22:35.440 align:start position:0% general. We're still so early in all of 00:22:35.440 --> 00:22:37.510 align:start position:0% general. We're still so early in all of this.<00:22:35.840> Nobody<00:22:36.240> knows<00:22:36.480> what<00:22:36.640> we<00:22:36.880> can<00:22:37.039> and<00:22:37.280> can't 00:22:37.510 --> 00:22:37.520 align:start position:0% this. Nobody knows what we can and can't 00:22:37.520 --> 00:22:39.990 align:start position:0% this. Nobody knows what we can and can't do<00:22:38.240> like<00:22:38.559> technically<00:22:39.039> speaking,<00:22:39.440> much<00:22:39.679> less 00:22:39.990 --> 00:22:40.000 align:start position:0% do like technically speaking, much less 00:22:40.000 --> 00:22:42.230 align:start position:0% do like technically speaking, much less legally<00:22:40.480> speaking,<00:22:41.120> across<00:22:41.600> all<00:22:41.840> of<00:22:42.000> these 00:22:42.230 --> 00:22:42.240 align:start position:0% legally speaking, across all of these 00:22:42.240 --> 00:22:44.070 align:start position:0% legally speaking, across all of these things.<00:22:42.799> This<00:22:43.039> is<00:22:43.120> the<00:22:43.360> first<00:22:43.520> time<00:22:43.600> I've<00:22:43.919> seen 00:22:44.070 --> 00:22:44.080 align:start position:0% things. This is the first time I've seen 00:22:44.080 --> 00:22:46.549 align:start position:0% things. This is the first time I've seen someone<00:22:44.480> try<00:22:44.720> and<00:22:45.039> more<00:22:45.280> importantly<00:22:45.760> succeed 00:22:46.549 --> 00:22:46.559 align:start position:0% someone try and more importantly succeed 00:22:46.559 --> 00:22:49.190 align:start position:0% someone try and more importantly succeed in<00:22:47.120> skirting<00:22:47.520> the<00:22:47.760> license.<00:22:48.480> How<00:22:48.720> big<00:22:48.880> of<00:22:49.039> a 00:22:49.190 --> 00:22:49.200 align:start position:0% in skirting the license. How big of a 00:22:49.200 --> 00:22:51.430 align:start position:0% in skirting the license. How big of a deal<00:22:49.360> is<00:22:49.600> this<00:22:49.919> is<00:22:50.320> up<00:22:50.559> for<00:22:50.880> debate. 00:22:51.430 --> 00:22:51.440 align:start position:0% deal is this is up for debate. 00:22:51.440 --> 00:22:53.110 align:start position:0% deal is this is up for debate. Personally,<00:22:52.080> I<00:22:52.320> would<00:22:52.400> have<00:22:52.559> liked<00:22:52.799> at<00:22:52.960> the 00:22:53.110 --> 00:22:53.120 align:start position:0% Personally, I would have liked at the 00:22:53.120 --> 00:22:55.190 align:start position:0% Personally, I would have liked at the very<00:22:53.360> least<00:22:53.679> for<00:22:53.919> the<00:22:54.159> cursor<00:22:54.640> team<00:22:54.799> to<00:22:55.039> have 00:22:55.190 --> 00:22:55.200 align:start position:0% very least for the cursor team to have 00:22:55.200 --> 00:22:56.950 align:start position:0% very least for the cursor team to have come<00:22:55.280> out<00:22:55.440> and<00:22:55.679> said<00:22:55.840> that<00:22:56.080> their<00:22:56.320> new<00:22:56.480> model 00:22:56.950 --> 00:22:56.960 align:start position:0% come out and said that their new model 00:22:56.960 --> 00:22:59.909 align:start position:0% come out and said that their new model is<00:22:57.360> a<00:22:57.600> post-trained<00:22:58.559> refined<00:22:59.520> different 00:22:59.909 --> 00:22:59.919 align:start position:0% is a post-trained refined different 00:22:59.919 --> 00:23:02.870 align:start position:0% is a post-trained refined different version<00:23:00.559> of<00:23:00.799> Kim<00:23:01.200> K<00:23:01.360> 2.5<00:23:02.080> and<00:23:02.320> try<00:23:02.400> to<00:23:02.559> take<00:23:02.720> the 00:23:02.870 --> 00:23:02.880 align:start position:0% version of Kim K 2.5 and try to take the 00:23:02.880 --> 00:23:05.029 align:start position:0% version of Kim K 2.5 and try to take the opportunity<00:23:03.200> to<00:23:03.520> educate<00:23:04.000> us<00:23:04.480> because<00:23:04.720> people 00:23:05.029 --> 00:23:05.039 align:start position:0% opportunity to educate us because people 00:23:05.039 --> 00:23:07.029 align:start position:0% opportunity to educate us because people seem<00:23:05.280> to<00:23:05.360> think<00:23:05.679> that<00:23:06.000> this<00:23:06.159> is<00:23:06.400> them<00:23:06.640> just 00:23:07.029 --> 00:23:07.039 align:start position:0% seem to think that this is them just 00:23:07.039 --> 00:23:08.789 align:start position:0% seem to think that this is them just stealing<00:23:07.360> the<00:23:07.520> model<00:23:07.760> and<00:23:08.000> rebranding<00:23:08.480> it<00:23:08.559> as 00:23:08.789 --> 00:23:08.799 align:start position:0% stealing the model and rebranding it as 00:23:08.799 --> 00:23:10.630 align:start position:0% stealing the model and rebranding it as theirs<00:23:09.039> and<00:23:09.280> not<00:23:09.440> changing<00:23:09.760> much<00:23:10.000> about<00:23:10.240> it. 00:23:10.630 --> 00:23:10.640 align:start position:0% theirs and not changing much about it. 00:23:10.640 --> 00:23:12.310 align:start position:0% theirs and not changing much about it. It's<00:23:10.880> not<00:23:11.039> that.<00:23:11.360> This<00:23:11.520> is<00:23:11.679> them<00:23:11.840> working<00:23:12.080> very 00:23:12.310 --> 00:23:12.320 align:start position:0% It's not that. This is them working very 00:23:12.320 --> 00:23:13.990 align:start position:0% It's not that. This is them working very hard<00:23:12.400> to<00:23:12.640> make<00:23:12.799> the<00:23:12.960> best<00:23:13.120> possible<00:23:13.520> thing<00:23:13.760> in 00:23:13.990 --> 00:23:14.000 align:start position:0% hard to make the best possible thing in 00:23:14.000 --> 00:23:15.830 align:start position:0% hard to make the best possible thing in their<00:23:14.159> tools<00:23:14.720> using<00:23:15.039> an<00:23:15.200> insane<00:23:15.600> amount<00:23:15.760> of 00:23:15.830 --> 00:23:15.840 align:start position:0% their tools using an insane amount of 00:23:15.840 --> 00:23:17.830 align:start position:0% their tools using an insane amount of compute<00:23:16.159> and<00:23:16.400> research<00:23:16.799> and<00:23:16.960> time<00:23:17.440> and<00:23:17.679> then 00:23:17.830 --> 00:23:17.840 align:start position:0% compute and research and time and then 00:23:17.840 --> 00:23:19.750 align:start position:0% compute and research and time and then not<00:23:18.080> properly<00:23:18.559> disclosing<00:23:19.360> what<00:23:19.600> the 00:23:19.750 --> 00:23:19.760 align:start position:0% not properly disclosing what the 00:23:19.760 --> 00:23:22.710 align:start position:0% not properly disclosing what the starting<00:23:20.080> point<00:23:20.400> was.<00:23:21.200> I<00:23:21.520> don't<00:23:21.679> like<00:23:21.919> that.<00:23:22.480> I 00:23:22.710 --> 00:23:22.720 align:start position:0% starting point was. I don't like that. I 00:23:22.720 --> 00:23:24.549 align:start position:0% starting point was. I don't like that. I think<00:23:22.880> this<00:23:23.039> isn't<00:23:23.360> within<00:23:23.679> the<00:23:23.840> spirit<00:23:24.240> of 00:23:24.549 --> 00:23:24.559 align:start position:0% think this isn't within the spirit of 00:23:24.559 --> 00:23:26.470 align:start position:0% think this isn't within the spirit of openw<00:23:24.960> weight<00:23:25.200> models.<00:23:25.679> And<00:23:25.840> obviously<00:23:26.240> open 00:23:26.470 --> 00:23:26.480 align:start position:0% openw weight models. And obviously open 00:23:26.480 --> 00:23:27.990 align:start position:0% openw weight models. And obviously open weight<00:23:26.720> does<00:23:26.880> not<00:23:27.039> mean<00:23:27.200> open<00:23:27.440> source.<00:23:27.840> You 00:23:27.990 --> 00:23:28.000 align:start position:0% weight does not mean open source. You 00:23:28.000 --> 00:23:29.669 align:start position:0% weight does not mean open source. You don't<00:23:28.159> have<00:23:28.320> all<00:23:28.559> of<00:23:28.640> the<00:23:28.799> things<00:23:28.960> you<00:23:29.200> need<00:23:29.440> to 00:23:29.669 --> 00:23:29.679 align:start position:0% don't have all of the things you need to 00:23:29.679 --> 00:23:31.190 align:start position:0% don't have all of the things you need to recreate<00:23:30.080> the<00:23:30.240> model<00:23:30.480> from<00:23:30.640> scratch<00:23:30.880> yourself 00:23:31.190 --> 00:23:31.200 align:start position:0% recreate the model from scratch yourself 00:23:31.200 --> 00:23:32.950 align:start position:0% recreate the model from scratch yourself just<00:23:31.360> cuz<00:23:31.520> it's<00:23:31.679> open<00:23:31.919> weight.<00:23:32.480> But<00:23:32.720> open 00:23:32.950 --> 00:23:32.960 align:start position:0% just cuz it's open weight. But open 00:23:32.960 --> 00:23:34.310 align:start position:0% just cuz it's open weight. But open weight<00:23:33.200> is<00:23:33.360> open<00:23:33.600> weight.<00:23:33.840> It<00:23:34.000> means<00:23:34.159> the 00:23:34.310 --> 00:23:34.320 align:start position:0% weight is open weight. It means the 00:23:34.320 --> 00:23:35.510 align:start position:0% weight is open weight. It means the starting<00:23:34.559> point<00:23:34.720> is<00:23:34.880> out<00:23:35.039> there<00:23:35.120> for<00:23:35.280> people 00:23:35.510 --> 00:23:35.520 align:start position:0% starting point is out there for people 00:23:35.520 --> 00:23:37.590 align:start position:0% starting point is out there for people to<00:23:35.679> do<00:23:35.840> things<00:23:36.080> with.<00:23:36.559> And<00:23:36.799> the<00:23:37.120> whole<00:23:37.360> point 00:23:37.590 --> 00:23:37.600 align:start position:0% to do things with. And the whole point 00:23:37.600 --> 00:23:39.750 align:start position:0% to do things with. And the whole point of<00:23:37.919> doing<00:23:38.159> that<00:23:38.320> is<00:23:38.559> so<00:23:38.720> we<00:23:38.880> can<00:23:39.039> build<00:23:39.280> on<00:23:39.600> top 00:23:39.750 --> 00:23:39.760 align:start position:0% of doing that is so we can build on top 00:23:39.760 --> 00:23:42.470 align:start position:0% of doing that is so we can build on top of<00:23:39.840> it<00:23:40.240> and<00:23:40.559> do<00:23:40.720> cool<00:23:41.039> things<00:23:41.360> with<00:23:41.760> what<00:23:42.000> we 00:23:42.470 --> 00:23:42.480 align:start position:0% of it and do cool things with what we 00:23:42.480 --> 00:23:44.390 align:start position:0% of it and do cool things with what we have.<00:23:43.039> The<00:23:43.200> weird<00:23:43.600> part<00:23:43.760> here<00:23:44.000> is<00:23:44.159> that 00:23:44.390 --> 00:23:44.400 align:start position:0% have. The weird part here is that 00:23:44.400 --> 00:23:45.909 align:start position:0% have. The weird part here is that traditionally<00:23:44.880> with<00:23:45.120> open<00:23:45.360> source,<00:23:45.679> we 00:23:45.909 --> 00:23:45.919 align:start position:0% traditionally with open source, we 00:23:45.919 --> 00:23:47.750 align:start position:0% traditionally with open source, we expect<00:23:46.159> that<00:23:46.320> to<00:23:46.480> be<00:23:46.799> independent<00:23:47.360> groups, 00:23:47.750 --> 00:23:47.760 align:start position:0% expect that to be independent groups, 00:23:47.760 --> 00:23:49.590 align:start position:0% expect that to be independent groups, side<00:23:48.080> projects,<00:23:48.640> hobbyists,<00:23:49.280> and<00:23:49.440> people 00:23:49.590 --> 00:23:49.600 align:start position:0% side projects, hobbyists, and people 00:23:49.600 --> 00:23:51.510 align:start position:0% side projects, hobbyists, and people like<00:23:49.840> that<00:23:50.240> building<00:23:50.559> on<00:23:50.799> top.<00:23:51.120> But<00:23:51.280> all<00:23:51.360> you 00:23:51.510 --> 00:23:51.520 align:start position:0% like that building on top. But all you 00:23:51.520 --> 00:23:52.950 align:start position:0% like that building on top. But all you can<00:23:51.600> really<00:23:51.760> do<00:23:51.840> with<00:23:52.000> an<00:23:52.159> openweight<00:23:52.720> model 00:23:52.950 --> 00:23:52.960 align:start position:0% can really do with an openweight model 00:23:52.960 --> 00:23:55.430 align:start position:0% can really do with an openweight model is<00:23:53.360> run<00:23:53.520> it<00:23:53.679> on<00:23:53.840> your<00:23:53.919> own<00:23:54.080> GPUs<00:23:54.880> unless<00:23:55.200> you 00:23:55.430 --> 00:23:55.440 align:start position:0% is run it on your own GPUs unless you 00:23:55.440 --> 00:23:57.430 align:start position:0% is run it on your own GPUs unless you have<00:23:55.840> hundreds<00:23:56.240> of<00:23:56.400> millions<00:23:56.720> of<00:23:56.880> dollars<00:23:57.200> to 00:23:57.430 --> 00:23:57.440 align:start position:0% have hundreds of millions of dollars to 00:23:57.440 --> 00:23:59.110 align:start position:0% have hundreds of millions of dollars to spend<00:23:57.760> trying<00:23:58.159> to<00:23:58.320> adjust<00:23:58.640> it<00:23:58.799> to<00:23:58.960> do 00:23:59.110 --> 00:23:59.120 align:start position:0% spend trying to adjust it to do 00:23:59.120 --> 00:24:00.710 align:start position:0% spend trying to adjust it to do different<00:23:59.440> things,<00:23:59.919> which<00:24:00.080> is<00:24:00.159> what<00:24:00.320> cursor 00:24:00.710 --> 00:24:00.720 align:start position:0% different things, which is what cursor 00:24:00.720 --> 00:24:02.549 align:start position:0% different things, which is what cursor can<00:24:00.880> do.<00:24:01.039> So,<00:24:01.200> it's<00:24:01.440> one<00:24:01.520> of<00:24:01.600> the<00:24:01.760> few<00:24:02.000> times<00:24:02.320> we 00:24:02.549 --> 00:24:02.559 align:start position:0% can do. So, it's one of the few times we 00:24:02.559 --> 00:24:04.950 align:start position:0% can do. So, it's one of the few times we have<00:24:02.799> actually<00:24:03.120> had<00:24:03.760> a<00:24:04.000> company<00:24:04.480> that<00:24:04.720> could 00:24:04.950 --> 00:24:04.960 align:start position:0% have actually had a company that could 00:24:04.960 --> 00:24:07.110 align:start position:0% have actually had a company that could use<00:24:05.120> these<00:24:05.360> models<00:24:05.760> to<00:24:06.000> do<00:24:06.240> something<00:24:06.720> bigger 00:24:07.110 --> 00:24:07.120 align:start position:0% use these models to do something bigger 00:24:07.120 --> 00:24:09.830 align:start position:0% use these models to do something bigger and<00:24:07.360> bolder<00:24:08.000> than<00:24:08.480> we<00:24:08.720> had<00:24:08.880> seen<00:24:09.120> before.<00:24:09.600> I 00:24:09.830 --> 00:24:09.840 align:start position:0% and bolder than we had seen before. I 00:24:09.840 --> 00:24:10.870 align:start position:0% and bolder than we had seen before. I don't<00:24:10.000> think<00:24:10.080> there's<00:24:10.320> been<00:24:10.480> anything<00:24:10.640> like 00:24:10.870 --> 00:24:10.880 align:start position:0% don't think there's been anything like 00:24:10.880 --> 00:24:12.549 align:start position:0% don't think there's been anything like this<00:24:11.039> in<00:24:11.279> the<00:24:11.360> past,<00:24:11.760> but<00:24:11.919> I<00:24:12.080> do<00:24:12.240> think<00:24:12.320> there's 00:24:12.549 --> 00:24:12.559 align:start position:0% this in the past, but I do think there's 00:24:12.559 --> 00:24:14.149 align:start position:0% this in the past, but I do think there's going<00:24:12.640> to<00:24:12.720> be<00:24:12.799> a<00:24:12.960> lot<00:24:13.039> more<00:24:13.360> things<00:24:13.520> like<00:24:13.760> this 00:24:14.149 --> 00:24:14.159 align:start position:0% going to be a lot more things like this 00:24:14.159 --> 00:24:16.070 align:start position:0% going to be a lot more things like this going<00:24:14.480> forward.<00:24:15.039> As<00:24:15.440> companies<00:24:15.840> are 00:24:16.070 --> 00:24:16.080 align:start position:0% going forward. As companies are 00:24:16.080 --> 00:24:17.510 align:start position:0% going forward. As companies are integrating<00:24:16.480> these<00:24:16.720> models<00:24:17.039> and<00:24:17.279> they're 00:24:17.510 --> 00:24:17.520 align:start position:0% integrating these models and they're 00:24:17.520 --> 00:24:19.909 align:start position:0% integrating these models and they're trying<00:24:17.679> to<00:24:17.919> get<00:24:18.080> out<00:24:18.320> of<00:24:18.559> the<00:24:18.960> absurd<00:24:19.520> pricing 00:24:19.909 --> 00:24:19.919 align:start position:0% trying to get out of the absurd pricing 00:24:19.919 --> 00:24:21.830 align:start position:0% trying to get out of the absurd pricing subsidization<00:24:20.559> wars<00:24:21.039> that<00:24:21.279> companies<00:24:21.600> like 00:24:21.830 --> 00:24:21.840 align:start position:0% subsidization wars that companies like 00:24:21.840 --> 00:24:23.830 align:start position:0% subsidization wars that companies like Enthropic<00:24:22.400> and<00:24:22.480> OpenAI<00:24:23.039> are<00:24:23.200> doing,<00:24:23.679> they 00:24:23.830 --> 00:24:23.840 align:start position:0% Enthropic and OpenAI are doing, they 00:24:23.840 --> 00:24:26.230 align:start position:0% Enthropic and OpenAI are doing, they need<00:24:24.000> an<00:24:24.240> edge.<00:24:24.720> Openweight<00:24:25.279> models<00:24:25.760> are<00:24:26.000> that 00:24:26.230 --> 00:24:26.240 align:start position:0% need an edge. Openweight models are that 00:24:26.240 --> 00:24:28.390 align:start position:0% need an edge. Openweight models are that edge<00:24:26.640> in<00:24:26.799> a<00:24:27.039> lot<00:24:27.120> of<00:24:27.200> ways<00:24:27.600> because<00:24:27.840> the<00:24:28.080> prices 00:24:28.390 --> 00:24:28.400 align:start position:0% edge in a lot of ways because the prices 00:24:28.400 --> 00:24:30.470 align:start position:0% edge in a lot of ways because the prices are<00:24:28.640> just<00:24:28.960> absurdly<00:24:29.520> cheaper.<00:24:30.000> As<00:24:30.159> I<00:24:30.320> had 00:24:30.470 --> 00:24:30.480 align:start position:0% are just absurdly cheaper. As I had 00:24:30.480 --> 00:24:33.029 align:start position:0% are just absurdly cheaper. As I had mentioned<00:24:30.799> before,<00:24:31.440> Opus'<00:24:32.000> pricing<00:24:32.400> is 00:24:33.029 --> 00:24:33.039 align:start position:0% mentioned before, Opus' pricing is 00:24:33.039 --> 00:24:35.269 align:start position:0% mentioned before, Opus' pricing is expensive.<00:24:33.679> We're<00:24:33.919> talking<00:24:34.320> $5<00:24:34.799> per<00:24:35.039> million 00:24:35.269 --> 00:24:35.279 align:start position:0% expensive. We're talking $5 per million 00:24:35.279 --> 00:24:37.750 align:start position:0% expensive. We're talking $5 per million tokens<00:24:35.600> in<00:24:35.760> and<00:24:36.080> $25<00:24:36.720> per<00:24:36.960> million<00:24:37.279> out.<00:24:37.600> If 00:24:37.750 --> 00:24:37.760 align:start position:0% tokens in and $25 per million out. If 00:24:37.760 --> 00:24:40.549 align:start position:0% tokens in and $25 per million out. If you<00:24:37.919> look<00:24:38.000> at<00:24:38.159> Kimmy<00:24:38.559> K2.5,<00:24:39.600> it's<00:24:39.919> closer<00:24:40.159> to 00:24:40.549 --> 00:24:40.559 align:start position:0% you look at Kimmy K2.5, it's closer to 00:24:40.559 --> 00:24:42.630 align:start position:0% you look at Kimmy K2.5, it's closer to 45<00:24:40.960> cents<00:24:41.200> per<00:24:41.360> mill<00:24:41.520> in<00:24:41.679> and<00:24:41.840> 220<00:24:42.240> per<00:24:42.480> mill 00:24:42.630 --> 00:24:42.640 align:start position:0% 45 cents per mill in and 220 per mill 00:24:42.640 --> 00:24:44.390 align:start position:0% 45 cents per mill in and 220 per mill out.<00:24:42.960> So,<00:24:43.120> cursors<00:24:43.600> charging<00:24:43.919> a<00:24:44.159> little<00:24:44.240> bit 00:24:44.390 --> 00:24:44.400 align:start position:0% out. So, cursors charging a little bit 00:24:44.400 --> 00:24:46.149 align:start position:0% out. So, cursors charging a little bit more<00:24:44.559> than<00:24:44.640> that,<00:24:44.880> but<00:24:45.200> not<00:24:45.360> much.<00:24:45.840> Or<00:24:46.080> you 00:24:46.149 --> 00:24:46.159 align:start position:0% more than that, but not much. Or you 00:24:46.159 --> 00:24:49.110 align:start position:0% more than that, but not much. Or you look<00:24:46.240> at<00:24:46.480> models<00:24:46.720> like<00:24:46.960> GLM5,<00:24:48.240> 72<00:24:48.720> cents<00:24:48.960> per 00:24:49.110 --> 00:24:49.120 align:start position:0% look at models like GLM5, 72 cents per 00:24:49.120 --> 00:24:51.510 align:start position:0% look at models like GLM5, 72 cents per mill<00:24:49.279> in<00:24:49.440> and<00:24:49.679> 230<00:24:50.080> per<00:24:50.320> mill<00:24:50.559> out.<00:24:50.880> Huge<00:24:51.200> gap. 00:24:51.510 --> 00:24:51.520 align:start position:0% mill in and 230 per mill out. Huge gap. 00:24:51.520 --> 00:24:53.590 align:start position:0% mill in and 230 per mill out. Huge gap. These<00:24:51.760> are<00:24:52.000> 10x<00:24:52.480> or<00:24:52.720> more<00:24:52.960> cheaper.<00:24:53.520> And 00:24:53.590 --> 00:24:53.600 align:start position:0% These are 10x or more cheaper. And 00:24:53.600 --> 00:24:56.230 align:start position:0% These are 10x or more cheaper. And they're<00:24:53.840> not<00:24:54.000> 10x<00:24:54.480> dumber.<00:24:55.039> They're<00:24:55.279> like<00:24:55.679> 20% 00:24:56.230 --> 00:24:56.240 align:start position:0% they're not 10x dumber. They're like 20% 00:24:56.240 --> 00:24:59.190 align:start position:0% they're not 10x dumber. They're like 20% dumber<00:24:56.720> for<00:24:57.039> 20x<00:24:57.760> cheaper.<00:24:58.480> That's<00:24:58.720> insane. 00:24:59.190 --> 00:24:59.200 align:start position:0% dumber for 20x cheaper. That's insane. 00:24:59.200 --> 00:25:01.430 align:start position:0% dumber for 20x cheaper. That's insane. So<00:24:59.440> if<00:24:59.600> you're<00:24:59.760> trying<00:25:00.000> to<00:25:00.159> deal<00:25:00.320> with<00:25:00.480> the<00:25:00.799> 25x 00:25:01.430 --> 00:25:01.440 align:start position:0% So if you're trying to deal with the 25x 00:25:01.440 --> 00:25:03.350 align:start position:0% So if you're trying to deal with the 25x subsidization<00:25:02.240> that<00:25:02.480> is<00:25:02.720> being<00:25:02.880> done<00:25:03.120> by 00:25:03.350 --> 00:25:03.360 align:start position:0% subsidization that is being done by 00:25:03.360 --> 00:25:05.269 align:start position:0% subsidization that is being done by companies<00:25:03.679> like<00:25:04.000> Anthropic<00:25:04.720> where<00:25:04.880> you<00:25:05.039> spend 00:25:05.269 --> 00:25:05.279 align:start position:0% companies like Anthropic where you spend 00:25:05.279 --> 00:25:07.909 align:start position:0% companies like Anthropic where you spend 200<00:25:05.600> and<00:25:05.760> get<00:25:06.000> 5<00:25:06.240> grand,<00:25:06.880> well<00:25:07.200> if<00:25:07.440> you<00:25:07.600> find<00:25:07.760> a 00:25:07.909 --> 00:25:07.919 align:start position:0% 200 and get 5 grand, well if you find a 00:25:07.919 --> 00:25:10.310 align:start position:0% 200 and get 5 grand, well if you find a model<00:25:08.240> that's<00:25:08.720> way<00:25:09.039> cheaper,<00:25:09.440> 10<00:25:09.679> to<00:25:09.919> 20<00:25:10.080> times 00:25:10.310 --> 00:25:10.320 align:start position:0% model that's way cheaper, 10 to 20 times 00:25:10.320 --> 00:25:12.710 align:start position:0% model that's way cheaper, 10 to 20 times cheaper,<00:25:10.960> you<00:25:11.120> can<00:25:11.520> cover<00:25:11.919> a<00:25:12.240> lot<00:25:12.320> of<00:25:12.480> that 00:25:12.710 --> 00:25:12.720 align:start position:0% cheaper, you can cover a lot of that 00:25:12.720 --> 00:25:14.789 align:start position:0% cheaper, you can cover a lot of that subsidization<00:25:13.440> gap.<00:25:13.919> But<00:25:14.000> those<00:25:14.240> models<00:25:14.559> are 00:25:14.789 --> 00:25:14.799 align:start position:0% subsidization gap. But those models are 00:25:14.799 --> 00:25:16.710 align:start position:0% subsidization gap. But those models are not<00:25:14.880> as<00:25:15.120> good<00:25:15.360> at<00:25:15.600> a<00:25:15.840> lot<00:25:16.000> of<00:25:16.159> things,<00:25:16.480> in 00:25:16.710 --> 00:25:16.720 align:start position:0% not as good at a lot of things, in 00:25:16.720 --> 00:25:18.789 align:start position:0% not as good at a lot of things, in particular<00:25:17.440> coding.<00:25:18.000> So<00:25:18.159> what<00:25:18.320> if<00:25:18.480> you<00:25:18.640> were 00:25:18.789 --> 00:25:18.799 align:start position:0% particular coding. So what if you were 00:25:18.799 --> 00:25:21.029 align:start position:0% particular coding. So what if you were willing<00:25:19.039> to<00:25:19.279> compromise<00:25:19.679> on<00:25:20.080> everything<00:25:20.480> else 00:25:21.029 --> 00:25:21.039 align:start position:0% willing to compromise on everything else 00:25:21.039 --> 00:25:22.630 align:start position:0% willing to compromise on everything else just<00:25:21.279> to<00:25:21.520> focus<00:25:21.760> on<00:25:21.840> the<00:25:22.000> coding<00:25:22.320> side<00:25:22.400> and 00:25:22.630 --> 00:25:22.640 align:start position:0% just to focus on the coding side and 00:25:22.640 --> 00:25:23.750 align:start position:0% just to focus on the coding side and make<00:25:22.799> something<00:25:22.960> that<00:25:23.120> was<00:25:23.279> really<00:25:23.440> good<00:25:23.600> at 00:25:23.750 --> 00:25:23.760 align:start position:0% make something that was really good at 00:25:23.760 --> 00:25:25.669 align:start position:0% make something that was really good at that?<00:25:24.080> And<00:25:24.240> that's<00:25:24.480> what<00:25:24.640> we<00:25:24.799> now<00:25:25.039> have<00:25:25.360> with 00:25:25.669 --> 00:25:25.679 align:start position:0% that? And that's what we now have with 00:25:25.679 --> 00:25:28.149 align:start position:0% that? And that's what we now have with Composer,<00:25:26.240> the<00:25:26.480> pretty<00:25:26.720> dang<00:25:27.120> cool,<00:25:27.600> powerful 00:25:28.149 --> 00:25:28.159 align:start position:0% Composer, the pretty dang cool, powerful 00:25:28.159 --> 00:25:30.630 align:start position:0% Composer, the pretty dang cool, powerful model.<00:25:28.799> I<00:25:29.039> just<00:25:29.279> wish<00:25:29.520> they<00:25:29.840> had<00:25:30.000> shared<00:25:30.320> a<00:25:30.480> bit 00:25:30.630 --> 00:25:30.640 align:start position:0% model. I just wish they had shared a bit 00:25:30.640 --> 00:25:32.789 align:start position:0% model. I just wish they had shared a bit more<00:25:30.880> and<00:25:31.120> took<00:25:31.279> the<00:25:31.520> opportunity<00:25:32.000> to<00:25:32.240> educate 00:25:32.789 --> 00:25:32.799 align:start position:0% more and took the opportunity to educate 00:25:32.799 --> 00:25:34.710 align:start position:0% more and took the opportunity to educate us<00:25:33.200> about<00:25:33.520> what<00:25:33.679> this<00:25:33.919> type<00:25:34.080> of<00:25:34.159> post<00:25:34.400> training 00:25:34.710 --> 00:25:34.720 align:start position:0% us about what this type of post training 00:25:34.720 --> 00:25:36.549 align:start position:0% us about what this type of post training looks<00:25:34.880> like<00:25:35.120> instead<00:25:35.360> of<00:25:35.440> us<00:25:35.679> having<00:25:35.919> to<00:25:36.320> start 00:25:36.549 --> 00:25:36.559 align:start position:0% looks like instead of us having to start 00:25:36.559 --> 00:25:38.230 align:start position:0% looks like instead of us having to start a<00:25:36.720> whole<00:25:36.880> bunch<00:25:36.960> of<00:25:37.120> drama,<00:25:37.520> flame<00:25:37.919> my<00:25:38.080> and 00:25:38.230 --> 00:25:38.240 align:start position:0% a whole bunch of drama, flame my and 00:25:38.240 --> 00:25:40.149 align:start position:0% a whole bunch of drama, flame my and everyone<00:25:38.480> else's<00:25:38.799> DMs<00:25:39.120> in<00:25:39.279> the<00:25:39.440> process.<00:25:39.840> And 00:25:40.149 --> 00:25:40.159 align:start position:0% everyone else's DMs in the process. And 00:25:40.159 --> 00:25:41.750 align:start position:0% everyone else's DMs in the process. And now<00:25:40.400> I'm<00:25:40.559> here<00:25:40.799> filming<00:25:41.039> on<00:25:41.200> a<00:25:41.360> Saturday 00:25:41.750 --> 00:25:41.760 align:start position:0% now I'm here filming on a Saturday 00:25:41.760 --> 00:25:43.590 align:start position:0% now I'm here filming on a Saturday evening<00:25:42.400> just<00:25:42.559> to<00:25:42.799> make<00:25:42.960> sure<00:25:43.120> we<00:25:43.360> all 00:25:43.590 --> 00:25:43.600 align:start position:0% evening just to make sure we all 00:25:43.600 --> 00:25:45.269 align:start position:0% evening just to make sure we all understand<00:25:44.000> what<00:25:44.320> happened.<00:25:44.720> There<00:25:44.960> is<00:25:45.120> one 00:25:45.269 --> 00:25:45.279 align:start position:0% understand what happened. There is one 00:25:45.279 --> 00:25:46.630 align:start position:0% understand what happened. There is one last<00:25:45.520> thing<00:25:45.760> here<00:25:45.919> I<00:25:46.159> think<00:25:46.240> is<00:25:46.480> worth 00:25:46.630 --> 00:25:46.640 align:start position:0% last thing here I think is worth 00:25:46.640 --> 00:25:49.190 align:start position:0% last thing here I think is worth covering.<00:25:47.360> The<00:25:47.600> long-term<00:25:48.159> impact.<00:25:48.799> Dax 00:25:49.190 --> 00:25:49.200 align:start position:0% covering. The long-term impact. Dax 00:25:49.200 --> 00:25:50.789 align:start position:0% covering. The long-term impact. Dax posted<00:25:49.520> this<00:25:49.679> before<00:25:50.000> things<00:25:50.240> had<00:25:50.480> settled 00:25:50.789 --> 00:25:50.799 align:start position:0% posted this before things had settled 00:25:50.799 --> 00:25:52.149 align:start position:0% posted this before things had settled down,<00:25:50.960> but<00:25:51.120> I<00:25:51.279> think<00:25:51.360> it's<00:25:51.520> as<00:25:51.760> relevant<00:25:52.000> now 00:25:52.149 --> 00:25:52.159 align:start position:0% down, but I think it's as relevant now 00:25:52.159 --> 00:25:54.789 align:start position:0% down, but I think it's as relevant now as<00:25:52.400> ever.<00:25:52.720> Whether<00:25:53.279> this<00:25:53.679> being<00:25:53.919> the<00:25:54.240> drama 00:25:54.789 --> 00:25:54.799 align:start position:0% as ever. Whether this being the drama 00:25:54.799 --> 00:25:57.510 align:start position:0% as ever. Whether this being the drama around<00:25:55.200> using<00:25:55.679> Kimmy<00:25:56.080> K2.5<00:25:56.720> is<00:25:56.960> true<00:25:57.120> or<00:25:57.279> not, 00:25:57.510 --> 00:25:57.520 align:start position:0% around using Kimmy K2.5 is true or not, 00:25:57.520 --> 00:25:59.190 align:start position:0% around using Kimmy K2.5 is true or not, it's<00:25:57.760> going<00:25:57.919> to<00:25:58.080> cause<00:25:58.400> every<00:25:58.720> company 00:25:59.190 --> 00:25:59.200 align:start position:0% it's going to cause every company 00:25:59.200 --> 00:26:00.950 align:start position:0% it's going to cause every company producing<00:25:59.679> open-<00:26:00.000> source<00:26:00.240> models<00:26:00.640> to 00:26:00.950 --> 00:26:00.960 align:start position:0% producing open- source models to 00:26:00.960 --> 00:26:02.870 align:start position:0% producing open- source models to re-evaluate<00:26:01.600> if<00:26:01.840> they<00:26:02.000> should<00:26:02.240> continue<00:26:02.640> to 00:26:02.870 --> 00:26:02.880 align:start position:0% re-evaluate if they should continue to 00:26:02.880 --> 00:26:05.110 align:start position:0% re-evaluate if they should continue to do<00:26:03.039> so.<00:26:03.360> That<00:26:03.600> is<00:26:03.679> incredibly<00:26:04.159> frustrating.<00:26:04.880> I 00:26:05.110 --> 00:26:05.120 align:start position:0% do so. That is incredibly frustrating. I 00:26:05.120 --> 00:26:07.269 align:start position:0% do so. That is incredibly frustrating. I agree.<00:26:05.600> There's<00:26:05.840> a<00:26:06.000> very<00:26:06.240> good<00:26:06.400> chance<00:26:06.960> that 00:26:07.269 --> 00:26:07.279 align:start position:0% agree. There's a very good chance that 00:26:07.279 --> 00:26:08.870 align:start position:0% agree. There's a very good chance that Cursor<00:26:07.760> ends<00:26:07.919> up<00:26:08.080> making<00:26:08.320> more<00:26:08.559> money<00:26:08.720> on 00:26:08.870 --> 00:26:08.880 align:start position:0% Cursor ends up making more money on 00:26:08.880 --> 00:26:11.990 align:start position:0% Cursor ends up making more money on Kimmy<00:26:09.360> K2.5<00:26:10.240> than<00:26:10.480> Moonshot<00:26:11.039> does.<00:26:11.440> Hell, 00:26:11.990 --> 00:26:12.000 align:start position:0% Kimmy K2.5 than Moonshot does. Hell, 00:26:12.000 --> 00:26:13.190 align:start position:0% Kimmy K2.5 than Moonshot does. Hell, there's<00:26:12.159> a<00:26:12.320> good<00:26:12.480> chance<00:26:12.640> that<00:26:12.720> they<00:26:12.960> spent 00:26:13.190 --> 00:26:13.200 align:start position:0% there's a good chance that they spent 00:26:13.200 --> 00:26:15.430 align:start position:0% there's a good chance that they spent more<00:26:13.520> money<00:26:14.080> doing<00:26:14.320> this<00:26:14.559> post<00:26:14.880> training<00:26:15.200> on 00:26:15.430 --> 00:26:15.440 align:start position:0% more money doing this post training on 00:26:15.440 --> 00:26:19.029 align:start position:0% more money doing this post training on Kimmy<00:26:15.840> K2.5<00:26:17.120> than<00:26:17.760> Moonshot<00:26:18.240> has<00:26:18.480> had<00:26:18.640> in<00:26:18.880> the 00:26:19.029 --> 00:26:19.039 align:start position:0% Kimmy K2.5 than Moonshot has had in the 00:26:19.039 --> 00:26:21.110 align:start position:0% Kimmy K2.5 than Moonshot has had in the entire<00:26:19.360> history<00:26:19.600> of<00:26:19.840> their<00:26:20.159> business.<00:26:20.799> That 00:26:21.110 --> 00:26:21.120 align:start position:0% entire history of their business. That 00:26:21.120 --> 00:26:22.789 align:start position:0% entire history of their business. That sucks.<00:26:21.679> And<00:26:21.840> if<00:26:22.000> you're<00:26:22.080> one<00:26:22.240> of<00:26:22.320> those<00:26:22.559> small 00:26:22.789 --> 00:26:22.799 align:start position:0% sucks. And if you're one of those small 00:26:22.799 --> 00:26:24.870 align:start position:0% sucks. And if you're one of those small labs<00:26:23.360> that<00:26:23.600> is<00:26:23.919> fighting<00:26:24.159> to<00:26:24.400> make<00:26:24.480> these 00:26:24.870 --> 00:26:24.880 align:start position:0% labs that is fighting to make these 00:26:24.880 --> 00:26:26.390 align:start position:0% labs that is fighting to make these powerful<00:26:25.279> models<00:26:25.600> and<00:26:25.760> the<00:26:25.919> only<00:26:26.080> way<00:26:26.159> you<00:26:26.320> can 00:26:26.390 --> 00:26:26.400 align:start position:0% powerful models and the only way you can 00:26:26.400 --> 00:26:27.750 align:start position:0% powerful models and the only way you can get<00:26:26.559> people<00:26:26.799> interested<00:26:27.120> is<00:26:27.360> releasing<00:26:27.600> them 00:26:27.750 --> 00:26:27.760 align:start position:0% get people interested is releasing them 00:26:27.760 --> 00:26:29.430 align:start position:0% get people interested is releasing them open<00:26:28.080> weight,<00:26:28.640> you're<00:26:28.799> now<00:26:28.960> looking<00:26:29.200> over 00:26:29.430 --> 00:26:29.440 align:start position:0% open weight, you're now looking over 00:26:29.440 --> 00:26:31.110 align:start position:0% open weight, you're now looking over your<00:26:29.600> shoulder<00:26:30.000> like<00:26:30.400> should<00:26:30.640> we<00:26:30.799> even<00:26:30.960> be 00:26:31.110 --> 00:26:31.120 align:start position:0% your shoulder like should we even be 00:26:31.120 --> 00:26:32.950 align:start position:0% your shoulder like should we even be doing<00:26:31.279> this?<00:26:31.840> And<00:26:32.000> that's<00:26:32.159> a<00:26:32.400> scary<00:26:32.720> thing 00:26:32.950 --> 00:26:32.960 align:start position:0% doing this? And that's a scary thing 00:26:32.960 --> 00:26:34.870 align:start position:0% doing this? And that's a scary thing because<00:26:33.279> we<00:26:33.520> need<00:26:33.679> these<00:26:33.919> openweight<00:26:34.480> models 00:26:34.870 --> 00:26:34.880 align:start position:0% because we need these openweight models 00:26:34.880 --> 00:26:37.669 align:start position:0% because we need these openweight models in<00:26:35.200> order<00:26:35.360> to<00:26:36.080> keep<00:26:36.320> things<00:26:36.640> progressing.<00:26:37.520> All 00:26:37.669 --> 00:26:37.679 align:start position:0% in order to keep things progressing. All 00:26:37.679 --> 00:26:39.269 align:start position:0% in order to keep things progressing. All the<00:26:37.919> cost<00:26:38.080> improvements<00:26:38.480> we've<00:26:38.720> seen<00:26:38.960> across 00:26:39.269 --> 00:26:39.279 align:start position:0% the cost improvements we've seen across 00:26:39.279 --> 00:26:41.029 align:start position:0% the cost improvements we've seen across the<00:26:39.520> flagship<00:26:40.000> models<00:26:40.320> are<00:26:40.559> because<00:26:40.799> the 00:26:41.029 --> 00:26:41.039 align:start position:0% the flagship models are because the 00:26:41.039 --> 00:26:42.789 align:start position:0% the flagship models are because the openweight<00:26:41.440> models<00:26:41.760> got<00:26:42.000> cheaper<00:26:42.240> and<00:26:42.480> better 00:26:42.789 --> 00:26:42.799 align:start position:0% openweight models got cheaper and better 00:26:42.799 --> 00:26:45.269 align:start position:0% openweight models got cheaper and better enough<00:26:43.120> that<00:26:43.440> these<00:26:43.760> companies<00:26:44.480> have<00:26:44.640> to 00:26:45.269 --> 00:26:45.279 align:start position:0% enough that these companies have to 00:26:45.279 --> 00:26:47.029 align:start position:0% enough that these companies have to rethink<00:26:45.679> things<00:26:46.159> and<00:26:46.320> I<00:26:46.480> really<00:26:46.640> hope<00:26:46.799> we 00:26:47.029 --> 00:26:47.039 align:start position:0% rethink things and I really hope we 00:26:47.039 --> 00:26:48.950 align:start position:0% rethink things and I really hope we don't<00:26:47.200> lose<00:26:47.440> that<00:26:47.679> as<00:26:47.840> a<00:26:48.080> result<00:26:48.320> of<00:26:48.480> how<00:26:48.720> this 00:26:48.950 --> 00:26:48.960 align:start position:0% don't lose that as a result of how this 00:26:48.960 --> 00:26:51.110 align:start position:0% don't lose that as a result of how this release<00:26:49.279> went.<00:26:49.679> So<00:26:49.919> wrapping<00:26:50.240> up,<00:26:50.640> Composer 00:26:51.110 --> 00:26:51.120 align:start position:0% release went. So wrapping up, Composer 00:26:51.120 --> 00:26:53.269 align:start position:0% release went. So wrapping up, Composer is<00:26:51.200> a<00:26:51.360> good<00:26:51.520> model.<00:26:51.919> Kimmy<00:26:52.240> K25<00:26:52.720> is<00:26:52.880> also<00:26:53.039> a 00:26:53.269 --> 00:26:53.279 align:start position:0% is a good model. Kimmy K25 is also a 00:26:53.279 --> 00:26:54.630 align:start position:0% is a good model. Kimmy K25 is also a good<00:26:53.360> model.<00:26:53.760> They<00:26:53.919> are<00:26:54.080> very,<00:26:54.320> very 00:26:54.630 --> 00:26:54.640 align:start position:0% good model. They are very, very 00:26:54.640 --> 00:26:56.070 align:start position:0% good model. They are very, very different<00:26:54.880> models.<00:26:55.279> A<00:26:55.520> lot<00:26:55.600> of<00:26:55.679> money<00:26:55.919> was 00:26:56.070 --> 00:26:56.080 align:start position:0% different models. A lot of money was 00:26:56.080 --> 00:26:57.990 align:start position:0% different models. A lot of money was spent<00:26:56.400> in<00:26:56.559> order<00:26:56.720> to<00:26:56.880> train<00:26:57.120> Composer<00:26:57.600> 2.<00:26:57.840> And 00:26:57.990 --> 00:26:58.000 align:start position:0% spent in order to train Composer 2. And 00:26:58.000 --> 00:26:59.269 align:start position:0% spent in order to train Composer 2. And they<00:26:58.240> definitely<00:26:58.559> should<00:26:58.720> have<00:26:58.880> disclosed 00:26:59.269 --> 00:26:59.279 align:start position:0% they definitely should have disclosed 00:26:59.279 --> 00:27:01.190 align:start position:0% they definitely should have disclosed the<00:26:59.440> fact<00:26:59.600> that<00:26:59.840> Composer<00:27:00.320> 2<00:27:00.559> was<00:27:00.799> being<00:27:00.960> built 00:27:01.190 --> 00:27:01.200 align:start position:0% the fact that Composer 2 was being built 00:27:01.200 --> 00:27:04.630 align:start position:0% the fact that Composer 2 was being built on<00:27:01.440> top<00:27:01.520> of<00:27:01.600> Kimmy<00:27:02.080> K2.5.<00:27:03.200> Everyone<00:27:04.240> kind<00:27:04.480> of 00:27:04.630 --> 00:27:04.640 align:start position:0% on top of Kimmy K2.5. Everyone kind of 00:27:04.640 --> 00:27:06.470 align:start position:0% on top of Kimmy K2.5. Everyone kind of screwed<00:27:05.039> up<00:27:05.279> here.<00:27:05.600> I<00:27:05.840> think<00:27:05.919> the<00:27:06.159> license 00:27:06.470 --> 00:27:06.480 align:start position:0% screwed up here. I think the license 00:27:06.480 --> 00:27:08.149 align:start position:0% screwed up here. I think the license terms<00:27:06.799> are<00:27:06.960> not<00:27:07.120> clear<00:27:07.360> enough<00:27:07.520> and<00:27:07.760> also<00:27:08.000> are 00:27:08.149 --> 00:27:08.159 align:start position:0% terms are not clear enough and also are 00:27:08.159 --> 00:27:09.669 align:start position:0% terms are not clear enough and also are not<00:27:08.320> well<00:27:08.480> defined<00:27:08.880> enough<00:27:09.039> for<00:27:09.200> us<00:27:09.360> to<00:27:09.520> know 00:27:09.669 --> 00:27:09.679 align:start position:0% not well defined enough for us to know 00:27:09.679 --> 00:27:11.669 align:start position:0% not well defined enough for us to know what<00:27:09.919> they<00:27:10.159> really<00:27:10.400> mean<00:27:10.640> in<00:27:10.880> practice.<00:27:11.440> I 00:27:11.669 --> 00:27:11.679 align:start position:0% what they really mean in practice. I 00:27:11.679 --> 00:27:12.950 align:start position:0% what they really mean in practice. I think<00:27:11.760> Fireworks<00:27:12.240> didn't<00:27:12.480> do<00:27:12.559> a<00:27:12.640> good<00:27:12.799> enough 00:27:12.950 --> 00:27:12.960 align:start position:0% think Fireworks didn't do a good enough 00:27:12.960 --> 00:27:14.390 align:start position:0% think Fireworks didn't do a good enough job<00:27:13.200> disclosing<00:27:13.760> and<00:27:13.919> forcing<00:27:14.240> the 00:27:14.390 --> 00:27:14.400 align:start position:0% job disclosing and forcing the 00:27:14.400 --> 00:27:16.230 align:start position:0% job disclosing and forcing the disclosure<00:27:14.799> to<00:27:14.960> go<00:27:15.120> down<00:27:15.279> the<00:27:15.440> pipeline.<00:27:16.000> And 00:27:16.230 --> 00:27:16.240 align:start position:0% disclosure to go down the pipeline. And 00:27:16.240 --> 00:27:17.669 align:start position:0% disclosure to go down the pipeline. And I<00:27:16.400> really<00:27:16.559> don't<00:27:16.720> think<00:27:16.880> Curser<00:27:17.360> did<00:27:17.440> a<00:27:17.600> great 00:27:17.669 --> 00:27:17.679 align:start position:0% I really don't think Curser did a great 00:27:17.679 --> 00:27:19.510 align:start position:0% I really don't think Curser did a great job<00:27:17.840> with<00:27:18.080> their<00:27:18.240> disclosures<00:27:18.799> here<00:27:19.120> either. 00:27:19.510 --> 00:27:19.520 align:start position:0% job with their disclosures here either. 00:27:19.520 --> 00:27:21.110 align:start position:0% job with their disclosures here either. I<00:27:19.760> am<00:27:19.840> disappointed<00:27:20.320> in<00:27:20.480> how<00:27:20.640> they<00:27:20.799> handled 00:27:21.110 --> 00:27:21.120 align:start position:0% I am disappointed in how they handled 00:27:21.120 --> 00:27:22.710 align:start position:0% I am disappointed in how they handled all<00:27:21.200> of<00:27:21.360> this,<00:27:21.520> but<00:27:21.679> I<00:27:21.840> am<00:27:22.000> thankful<00:27:22.240> that<00:27:22.480> they 00:27:22.710 --> 00:27:22.720 align:start position:0% all of this, but I am thankful that they 00:27:22.720 --> 00:27:24.070 align:start position:0% all of this, but I am thankful that they are<00:27:22.799> acknowledging<00:27:23.200> their<00:27:23.440> failures<00:27:23.760> and 00:27:24.070 --> 00:27:24.080 align:start position:0% are acknowledging their failures and 00:27:24.080 --> 00:27:25.750 align:start position:0% are acknowledging their failures and trying<00:27:24.240> to<00:27:24.400> do<00:27:24.480> better<00:27:24.720> in<00:27:24.960> the<00:27:25.120> future.<00:27:25.600> I 00:27:25.750 --> 00:27:25.760 align:start position:0% trying to do better in the future. I 00:27:25.760 --> 00:27:27.430 align:start position:0% trying to do better in the future. I hope,<00:27:26.080> fingers<00:27:26.480> crossed,<00:27:26.880> this<00:27:27.039> is<00:27:27.200> the<00:27:27.279> last 00:27:27.430 --> 00:27:27.440 align:start position:0% hope, fingers crossed, this is the last 00:27:27.440 --> 00:27:28.310 align:start position:0% hope, fingers crossed, this is the last time<00:27:27.520> we<00:27:27.679> see<00:27:27.840> something<00:27:28.000> like<00:27:28.080> this<00:27:28.159> from 00:27:28.310 --> 00:27:28.320 align:start position:0% time we see something like this from 00:27:28.320 --> 00:27:29.830 align:start position:0% time we see something like this from Cursor,<00:27:28.720> but<00:27:28.880> I<00:27:28.960> have<00:27:29.039> a<00:27:29.200> bad<00:27:29.360> feeling<00:27:29.600> that 00:27:29.830 --> 00:27:29.840 align:start position:0% Cursor, but I have a bad feeling that 00:27:29.840 --> 00:27:31.110 align:start position:0% Cursor, but I have a bad feeling that not<00:27:30.080> only<00:27:30.240> will<00:27:30.400> they<00:27:30.640> do<00:27:30.799> something<00:27:30.960> like 00:27:31.110 --> 00:27:31.120 align:start position:0% not only will they do something like 00:27:31.120 --> 00:27:32.789 align:start position:0% not only will they do something like this<00:27:31.279> again,<00:27:31.600> but<00:27:31.840> many<00:27:32.159> other<00:27:32.400> companies 00:27:32.789 --> 00:27:32.799 align:start position:0% this again, but many other companies 00:27:32.799 --> 00:27:34.630 align:start position:0% this again, but many other companies will<00:27:33.120> too.<00:27:33.440> We<00:27:33.600> are<00:27:33.760> now<00:27:33.840> in<00:27:34.000> an<00:27:34.240> era<00:27:34.400> where 00:27:34.630 --> 00:27:34.640 align:start position:0% will too. We are now in an era where 00:27:34.640 --> 00:27:36.470 align:start position:0% will too. We are now in an era where having<00:27:34.960> cheap<00:27:35.200> enough<00:27:35.600> compute<00:27:35.919> to<00:27:36.159> compete 00:27:36.470 --> 00:27:36.480 align:start position:0% having cheap enough compute to compete 00:27:36.480 --> 00:27:38.549 align:start position:0% having cheap enough compute to compete with<00:27:36.799> anthropic<00:27:37.200> and<00:27:37.360> open<00:27:37.600> AAI<00:27:38.159> really, 00:27:38.549 --> 00:27:38.559 align:start position:0% with anthropic and open AAI really, 00:27:38.559 --> 00:27:40.149 align:start position:0% with anthropic and open AAI really, really<00:27:38.960> matters.<00:27:39.520> And<00:27:39.600> the<00:27:39.679> way<00:27:39.919> companies 00:27:40.149 --> 00:27:40.159 align:start position:0% really matters. And the way companies 00:27:40.159 --> 00:27:41.110 align:start position:0% really matters. And the way companies get<00:27:40.320> there<00:27:40.480> are<00:27:40.640> going<00:27:40.640> to<00:27:40.799> get<00:27:40.880> more 00:27:41.110 --> 00:27:41.120 align:start position:0% get there are going to get more 00:27:41.120 --> 00:27:42.470 align:start position:0% get there are going to get more interesting<00:27:41.520> and<00:27:41.840> probably<00:27:42.159> more 00:27:42.470 --> 00:27:42.480 align:start position:0% interesting and probably more 00:27:42.480 --> 00:27:44.230 align:start position:0% interesting and probably more problematic<00:27:43.120> before<00:27:43.440> things<00:27:43.679> settle<00:27:44.000> down. 00:27:44.230 --> 00:27:44.240 align:start position:0% problematic before things settle down. 00:27:44.240 --> 00:27:47.360 align:start position:0% problematic before things settle down. Let<00:27:44.400> me<00:27:44.480> know<00:27:44.640> how<00:27:44.720> y'all<00:27:44.960> feel<00:27:45.120> about<00:27:45.200> this.