WEBVTT Kind: captions Language: en 00:00:01.955 --> 00:00:03.329 align:start position:0% Anthropic's <00:00:02.192>most <00:00:02.429>mysterious <00:00:02.666>model <00:00:02.903>has 00:00:03.329 --> 00:00:03.364 align:start position:0% Anthropic's most mysterious model has 00:00:03.364 --> 00:00:04.069 align:start position:0% Anthropic's most mysterious model has just 00:00:04.069 --> 00:00:04.105 align:start position:0% just 00:00:04.105 --> 00:00:05.479 align:start position:0% just leaked <00:00:04.375>ahead <00:00:04.645>of <00:00:04.915>time 00:00:05.479 --> 00:00:06.836 align:start position:0% leaked ahead of time . <00:00:05.686>Claude <00:00:05.893>Fable <00:00:06.100>5.5 <00:00:06.307>has <00:00:06.514>already 00:00:06.836 --> 00:00:06.889 align:start position:0% . Claude Fable 5.5 has already 00:00:06.889 --> 00:00:08.218 align:start position:0% . Claude Fable 5.5 has already appeared <00:00:07.526>in <00:00:08.163>Azure. 00:00:08.218 --> 00:00:08.318 align:start position:0% appeared in Azure. 00:00:08.318 --> 00:00:09.074 align:start position:0% appeared in Azure. The specifications 00:00:09.074 --> 00:00:09.101 align:start position:0% The specifications 00:00:09.101 --> 00:00:09.939 align:start position:0% The specifications have <00:00:09.371>been <00:00:09.641>leaked <00:00:09.911>online. 00:00:09.939 --> 00:00:10.039 align:start position:0% have been leaked online. 00:00:10.039 --> 00:00:10.901 align:start position:0% have been leaked online. Some <00:00:10.318>of <00:00:10.597>you <00:00:10.876>may 00:00:10.901 --> 00:00:10.925 align:start position:0% Some of you may 00:00:10.925 --> 00:00:11.834 align:start position:0% Some of you may already <00:00:11.220>be <00:00:11.515>using <00:00:11.810>it 00:00:11.834 --> 00:00:11.858 align:start position:0% already be using it 00:00:11.858 --> 00:00:12.721 align:start position:0% already be using it without <00:00:12.219>even <00:00:12.580>knowing 00:00:12.721 --> 00:00:12.744 align:start position:0% without even knowing 00:00:12.744 --> 00:00:13.807 align:start position:0% without even knowing it. <00:00:13.079>People <00:00:13.414>are <00:00:13.749>creating 00:00:13.807 --> 00:00:13.835 align:start position:0% it. People are creating 00:00:13.835 --> 00:00:14.880 align:start position:0% it. People are creating full-fledged <00:00:14.852>games, 00:00:14.880 --> 00:00:14.908 align:start position:0% full-fledged games, 00:00:14.908 --> 00:00:15.981 align:start position:0% full-fledged games, animated <00:00:15.388>films, <00:00:15.868>and 00:00:15.981 --> 00:00:16.009 align:start position:0% animated films, and 00:00:16.009 --> 00:00:16.856 align:start position:0% animated films, and even <00:00:16.281>symphonies <00:00:16.553>with <00:00:16.825>it. 00:00:16.856 --> 00:00:16.956 align:start position:0% even symphonies with it. 00:00:16.956 --> 00:00:18.424 align:start position:0% even symphonies with it. Some <00:00:17.342>believe <00:00:17.728>it <00:00:18.114>was 00:00:18.424 --> 00:00:18.463 align:start position:0% Some believe it was 00:00:18.463 --> 00:00:19.777 align:start position:0% Some believe it was created <00:00:18.695>by <00:00:18.927>AI, <00:00:19.159>and <00:00:19.391>rumors 00:00:19.777 --> 00:00:19.816 align:start position:0% created by AI, and rumors 00:00:19.816 --> 00:00:21.285 align:start position:0% created by AI, and rumors about <00:00:20.102>Opus <00:00:20.388>6 <00:00:20.674>are <00:00:20.960>already <00:00:21.246>spreading. 00:00:21.285 --> 00:00:21.385 align:start position:0% about Opus 6 are already spreading. 00:00:21.385 --> 00:00:22.849 align:start position:0% about Opus 6 are already spreading. In <00:00:21.610>addition, <00:00:21.835>Opus <00:00:22.060>5.5 <00:00:22.285>has <00:00:22.510>become 00:00:22.849 --> 00:00:22.891 align:start position:0% In addition, Opus 5.5 has become 00:00:22.891 --> 00:00:24.146 align:start position:0% In addition, Opus 5.5 has become much <00:00:23.477>faster <00:00:24.063>and 00:00:24.146 --> 00:00:24.188 align:start position:0% much faster and 00:00:24.188 --> 00:00:25.180 align:start position:0% much faster and cheaper. <00:00:25.000>Claude 00:00:25.180 --> 00:00:25.211 align:start position:0% cheaper. Claude 00:00:25.211 --> 00:00:25.844 align:start position:0% cheaper. Claude has <00:00:25.467>learned <00:00:25.723>not 00:00:25.844 --> 00:00:25.874 align:start position:0% has learned not 00:00:25.874 --> 00:00:26.689 align:start position:0% has learned not to <00:00:26.160>waste <00:00:26.446>your 00:00:26.689 --> 00:00:26.719 align:start position:0% to waste your 00:00:26.719 --> 00:00:27.891 align:start position:0% to waste your tokens. <00:00:27.664>OpenAI 00:00:27.891 --> 00:00:27.929 align:start position:0% tokens. OpenAI 00:00:27.929 --> 00:00:28.761 align:start position:0% tokens. OpenAI is <00:00:28.307>preparing <00:00:28.685>a 00:00:28.761 --> 00:00:28.799 align:start position:0% is preparing a 00:00:28.799 --> 00:00:29.851 align:start position:0% is preparing a counterattack. <00:00:29.258>And <00:00:29.717>Meta 00:00:29.851 --> 00:00:29.884 align:start position:0% counterattack. And Meta 00:00:29.884 --> 00:00:30.779 align:start position:0% counterattack. And Meta returned <00:00:30.647>to 00:00:30.779 --> 00:00:30.813 align:start position:0% returned to 00:00:30.813 --> 00:00:31.808 align:start position:0% returned to open <00:00:31.775>source, 00:00:31.808 --> 00:00:31.841 align:start position:0% open source, 00:00:31.841 --> 00:00:32.969 align:start position:0% open source, teaming <00:00:32.172>up, <00:00:32.503>oddly <00:00:32.834>enough 00:00:32.969 --> 00:00:33.002 align:start position:0% teaming up, oddly enough 00:00:33.002 --> 00:00:34.126 align:start position:0% teaming up, oddly enough , <00:00:33.365>with <00:00:33.728>Apple. <00:00:34.091>So, 00:00:34.126 --> 00:00:34.161 align:start position:0% , with Apple. So, 00:00:34.161 --> 00:00:35.512 align:start position:0% , with Apple. So, first <00:00:34.429>about <00:00:34.697>Fable, <00:00:34.965>because <00:00:35.233>no 00:00:35.512 --> 00:00:35.547 align:start position:0% first about Fable, because no 00:00:35.547 --> 00:00:36.586 align:start position:0% first about Fable, because no one <00:00:35.962>at <00:00:36.377>Anthropic 00:00:36.586 --> 00:00:36.621 align:start position:0% one at Anthropic 00:00:36.621 --> 00:00:37.903 align:start position:0% one at Anthropic announced <00:00:37.868>it. 00:00:37.903 --> 00:00:38.003 align:start position:0% announced it. 00:00:38.003 --> 00:00:38.981 align:start position:0% announced it. The <00:00:38.314>developers <00:00:38.625>simply 00:00:38.981 --> 00:00:39.011 align:start position:0% The developers simply 00:00:39.011 --> 00:00:39.782 align:start position:0% The developers simply found <00:00:39.248>it <00:00:39.485>in <00:00:39.722>the 00:00:39.782 --> 00:00:39.812 align:start position:0% found it in the 00:00:39.812 --> 00:00:40.761 align:start position:0% found it in the configuration 00:00:40.761 --> 00:00:40.790 align:start position:0% configuration 00:00:40.790 --> 00:00:41.903 align:start position:0% configuration file. <00:00:41.850>Code 00:00:41.903 --> 00:00:41.956 align:start position:0% file. Code 00:00:41.956 --> 00:00:43.877 align:start position:0% file. Code synced <00:00:42.783>by <00:00:43.610>Light 00:00:43.877 --> 00:00:43.931 align:start position:0% synced by Light 00:00:43.931 --> 00:00:45.852 align:start position:0% synced by Light LLM <00:00:44.233>showed <00:00:44.535>a <00:00:44.837>new <00:00:45.139>ID <00:00:45.441>in <00:00:45.743>the 00:00:45.852 --> 00:00:45.905 align:start position:0% LLM showed a new ID in the 00:00:45.905 --> 00:00:47.879 align:start position:0% LLM showed a new ID in the Azure 00:00:47.879 --> 00:00:47.933 align:start position:0% Azure 00:00:47.933 --> 00:00:50.014 align:start position:0% Azure East <00:00:48.199>US <00:00:48.465>2 <00:00:48.731>model <00:00:48.997>catalog, <00:00:49.263>azure.ai/claudefable55, <00:00:49.529>with <00:00:49.795>a 00:00:50.014 --> 00:00:50.067 align:start position:0% East US 2 model catalog, azure.ai/claudefable55, with a 00:00:50.067 --> 00:00:51.028 align:start position:0% East US 2 model catalog, azure.ai/claudefable55, with a status <00:00:50.520>of <00:00:50.973>" 00:00:51.028 --> 00:00:52.949 align:start position:0% status of " public." 00:00:52.949 --> 00:00:53.569 align:start position:0% public." 00:00:53.569 --> 00:00:56.014 align:start position:0% public." Remember <00:00:54.322>Fable <00:00:55.075>5.1? <00:00:55.828>It 00:00:56.014 --> 00:00:56.044 align:start position:0% Remember Fable 5.1? It 00:00:56.044 --> 00:00:56.965 align:start position:0% Remember Fable 5.1? It appeared <00:00:56.258>in <00:00:56.472>Bedrock, <00:00:56.686>and <00:00:56.900>a 00:00:56.965 --> 00:00:56.996 align:start position:0% appeared in Bedrock, and a 00:00:56.996 --> 00:00:57.855 align:start position:0% appeared in Bedrock, and a day <00:00:57.200>later <00:00:57.404>it <00:00:57.608>became 00:00:57.855 --> 00:00:57.885 align:start position:0% day later it became 00:00:57.885 --> 00:00:58.652 align:start position:0% day later it became available <00:00:58.468>to 00:00:58.652 --> 00:00:58.683 align:start position:0% available to 00:00:58.683 --> 00:00:59.235 align:start position:0% available to ordinary 00:00:59.235 --> 00:00:59.266 align:start position:0% ordinary 00:00:59.266 --> 00:01:00.402 align:start position:0% ordinary users <00:00:59.634>on <00:01:00.002>the <00:01:00.370>web. 00:01:00.402 --> 00:01:00.502 align:start position:0% users on the web. 00:01:00.502 --> 00:01:01.397 align:start position:0% users on the web. So <00:01:00.610>a <00:01:00.718>full 00:01:01.397 --> 00:01:01.427 align:start position:0% So a full 00:01:01.427 --> 00:01:02.539 align:start position:0% So a full release <00:01:01.658>is <00:01:01.889>probably <00:01:02.120>just <00:01:02.351>around 00:01:02.539 --> 00:01:02.570 align:start position:0% release is probably just around 00:01:02.570 --> 00:01:03.773 align:start position:0% release is probably just around the <00:01:03.156>corner, <00:01:03.742>maybe 00:01:03.773 --> 00:01:03.804 align:start position:0% the corner, maybe 00:01:03.804 --> 00:01:04.885 align:start position:0% the corner, maybe even <00:01:04.329>this <00:01:04.854>week. 00:01:04.885 --> 00:01:05.064 align:start position:0% even this week. 00:01:05.064 --> 00:01:06.108 align:start position:0% even this week. However, <00:01:05.516>not <00:01:05.968>everyone 00:01:06.108 --> 00:01:06.143 align:start position:0% However, not everyone 00:01:06.143 --> 00:01:07.292 align:start position:0% However, not everyone believes <00:01:06.291>this, <00:01:06.439>and <00:01:06.587>the <00:01:06.735>skeptics 00:01:07.292 --> 00:01:07.327 align:start position:0% believes this, and the skeptics 00:01:07.327 --> 00:01:09.285 align:start position:0% believes this, and the skeptics are <00:01:07.782>right. <00:01:08.237>Light <00:01:08.692>LLM <00:01:09.147>is 00:01:09.285 --> 00:01:09.320 align:start position:0% are right. Light LLM is 00:01:09.320 --> 00:01:10.395 align:start position:0% are right. Light LLM is open <00:01:09.719>source, <00:01:10.118>so 00:01:10.395 --> 00:01:10.430 align:start position:0% open source, so 00:01:10.430 --> 00:01:11.748 align:start position:0% open source, so its <00:01:10.846>synchronization <00:01:11.262>with <00:01:11.678>the 00:01:11.748 --> 00:01:11.783 align:start position:0% its synchronization with the 00:01:11.783 --> 00:01:13.066 align:start position:0% its synchronization with the Azure <00:01:12.072>catalog <00:01:12.361>is <00:01:12.650>completely 00:01:13.066 --> 00:01:13.101 align:start position:0% Azure catalog is completely 00:01:13.101 --> 00:01:14.211 align:start position:0% Azure catalog is completely normal, <00:01:13.343>and <00:01:13.585>it <00:01:13.827>is <00:01:14.069>not 00:01:14.211 --> 00:01:14.245 align:start position:0% normal, and it is not 00:01:14.245 --> 00:01:15.390 align:start position:0% normal, and it is not considered <00:01:14.609>a <00:01:14.973>second 00:01:15.390 --> 00:01:15.425 align:start position:0% considered a second 00:01:15.425 --> 00:01:16.743 align:start position:0% considered a second independent <00:01:16.708>source. 00:01:16.743 --> 00:01:16.843 align:start position:0% independent source. 00:01:16.843 --> 00:01:17.880 align:start position:0% independent source. And <00:01:16.949>the <00:01:17.055>fact <00:01:17.161>that <00:01:17.267>the <00:01:17.373>catalog 00:01:17.880 --> 00:01:17.911 align:start position:0% And the fact that the catalog 00:01:17.911 --> 00:01:19.042 align:start position:0% And the fact that the catalog indicates <00:01:18.112>the <00:01:18.313>GA <00:01:18.514>status <00:01:18.715>does <00:01:18.916>not 00:01:19.042 --> 00:01:19.074 align:start position:0% indicates the GA status does not 00:01:19.074 --> 00:01:19.702 align:start position:0% indicates the GA status does not mean <00:01:19.576>that 00:01:19.702 --> 00:01:19.734 align:start position:0% mean that 00:01:19.734 --> 00:01:20.645 align:start position:0% mean that the <00:01:19.898>announcement <00:01:20.062>will <00:01:20.226>be <00:01:20.390>made 00:01:20.645 --> 00:01:20.676 align:start position:0% the announcement will be made 00:01:20.676 --> 00:01:21.815 align:start position:0% the announcement will be made today. <00:01:21.166>Advice <00:01:21.656>from 00:01:21.815 --> 00:01:21.842 align:start position:0% today. Advice from 00:01:21.842 --> 00:01:22.240 align:start position:0% today. Advice from one <00:01:22.187>of 00:01:22.240 --> 00:01:22.267 align:start position:0% one of 00:01:22.267 --> 00:01:23.063 align:start position:0% one of the <00:01:22.612>users: <00:01:22.957>don't 00:01:23.063 --> 00:01:23.090 align:start position:0% the users: don't 00:01:23.090 --> 00:01:23.621 align:start position:0% the users: don't rush 00:01:23.621 --> 00:01:23.647 align:start position:0% rush 00:01:23.647 --> 00:01:24.471 align:start position:0% rush to <00:01:23.850>top <00:01:24.053>up <00:01:24.256>your 00:01:24.471 --> 00:01:24.497 align:start position:0% to top up your 00:01:24.497 --> 00:01:25.427 align:start position:0% to top up your account. <00:01:24.720>The <00:01:24.943>technical 00:01:25.427 --> 00:01:25.458 align:start position:0% account. The technical 00:01:25.458 --> 00:01:26.519 align:start position:0% account. The technical specifications <00:01:25.897>have <00:01:26.336>also 00:01:26.519 --> 00:01:26.549 align:start position:0% specifications have also 00:01:26.549 --> 00:01:27.897 align:start position:0% specifications have also been <00:01:26.826>leaked. <00:01:27.103>Context 00:01:27.897 --> 00:01:27.936 align:start position:0% been leaked. Context 00:01:27.936 --> 00:01:29.205 align:start position:0% been leaked. Context window <00:01:28.174>for <00:01:28.412>1 <00:01:28.650>million 00:01:29.205 --> 00:01:29.245 align:start position:0% window for 1 million 00:01:29.245 --> 00:01:30.355 align:start position:0% window for 1 million tokens, <00:01:29.562>128,000 <00:01:29.879>on <00:01:30.196>the 00:01:30.355 --> 00:01:30.395 align:start position:0% tokens, 128,000 on the 00:01:30.395 --> 00:01:31.703 align:start position:0% tokens, 128,000 on the way <00:01:30.640>out <00:01:30.885>and <00:01:31.130>API <00:01:31.375>price <00:01:31.620>of 00:01:31.703 --> 00:01:31.743 align:start position:0% way out and API price of 00:01:31.743 --> 00:01:33.091 align:start position:0% way out and API price of $10 <00:01:32.139>per <00:01:32.535>million 00:01:33.091 --> 00:01:33.131 align:start position:0% $10 per million 00:01:33.131 --> 00:01:34.519 align:start position:0% $10 per million inbound <00:01:33.785>tokens <00:01:34.439>and 00:01:34.519 --> 00:01:34.558 align:start position:0% inbound tokens and 00:01:34.558 --> 00:01:35.907 align:start position:0% inbound tokens and $50 <00:01:34.955>per <00:01:35.352>million 00:01:35.907 --> 00:01:35.946 align:start position:0% $50 per million 00:01:35.946 --> 00:01:37.249 align:start position:0% $50 per million outbound. <00:01:36.260>To <00:01:36.574>be <00:01:36.888>honest 00:01:37.249 --> 00:01:37.285 align:start position:0% outbound. To be honest 00:01:37.285 --> 00:01:38.581 align:start position:0% outbound. To be honest , 00:01:38.581 --> 00:01:38.617 align:start position:0% , 00:01:38.617 --> 00:01:39.949 align:start position:0% , these <00:01:38.876>numbers <00:01:39.135>themselves <00:01:39.394>are <00:01:39.653>not <00:01:39.912>new. 00:01:39.949 --> 00:01:40.238 align:start position:0% these numbers themselves are not new. 00:01:40.238 --> 00:01:41.848 align:start position:0% these numbers themselves are not new. Gemini <00:01:40.506>had <00:01:40.774>a <00:01:41.042>million 00:01:41.848 --> 00:01:41.905 align:start position:0% Gemini had a million 00:01:41.905 --> 00:01:44.091 align:start position:0% Gemini had a million context <00:01:42.883>tokens <00:01:43.861>back 00:01:44.091 --> 00:01:44.148 align:start position:0% context tokens back 00:01:44.148 --> 00:01:46.046 align:start position:0% context tokens back in <00:01:44.458>2024, <00:01:44.768>and <00:01:45.078>GPT-5 <00:01:45.388>was <00:01:45.698>already 00:01:46.046 --> 00:01:46.104 align:start position:0% in 2024, and GPT-5 was already 00:01:46.104 --> 00:01:48.346 align:start position:0% in 2024, and GPT-5 was already issuing <00:01:46.765>128,000 <00:01:47.426>last 00:01:48.346 --> 00:01:48.404 align:start position:0% issuing 128,000 last 00:01:48.404 --> 00:01:51.136 align:start position:0% issuing 128,000 last year. <00:01:49.713>However, <00:01:51.022>128,000 00:01:51.136 --> 00:01:51.173 align:start position:0% year. However, 128,000 00:01:51.173 --> 00:01:52.227 align:start position:0% year. However, 128,000 tokens <00:01:51.511>in <00:01:51.849>one 00:01:52.227 --> 00:01:52.264 align:start position:0% tokens in one 00:01:52.264 --> 00:01:53.205 align:start position:0% tokens in one answer <00:01:53.054>is 00:01:53.205 --> 00:01:53.242 align:start position:0% answer is 00:01:53.242 --> 00:01:54.710 align:start position:0% answer is effectively <00:01:53.562>a <00:01:53.882>full-fledged 00:01:54.710 --> 00:01:54.747 align:start position:0% effectively a full-fledged 00:01:54.747 --> 00:01:55.989 align:start position:0% effectively a full-fledged novel <00:01:54.916>or <00:01:55.085>the <00:01:55.254>entire <00:01:55.423>code <00:01:55.592>of <00:01:55.761>an 00:01:55.989 --> 00:01:56.026 align:start position:0% novel or the entire code of an 00:01:56.026 --> 00:01:57.381 align:start position:0% novel or the entire code of an average <00:01:57.343>project. 00:01:57.381 --> 00:01:58.621 align:start position:0% average project. 00:01:58.621 --> 00:01:59.780 align:start position:0% So <00:01:58.910>long <00:01:59.199>documents 00:01:59.780 --> 00:01:59.812 align:start position:0% So long documents 00:01:59.812 --> 00:02:01.036 align:start position:0% So long documents and <00:02:00.359>complex <00:02:00.906>agents 00:02:01.036 --> 00:02:01.069 align:start position:0% and complex agents 00:02:01.069 --> 00:02:02.325 align:start position:0% and complex agents working <00:02:01.616>continuously 00:02:02.325 --> 00:02:03.549 align:start position:0% working continuously are <00:02:02.808>exactly <00:02:03.291>what 00:02:03.549 --> 00:02:03.581 align:start position:0% are exactly what 00:02:03.581 --> 00:02:05.157 align:start position:0% are exactly what it <00:02:03.742>was <00:02:03.903>created <00:02:04.064>for. <00:02:04.225>The <00:02:04.386>catch <00:02:04.547>is <00:02:04.708>the 00:02:05.157 --> 00:02:05.194 align:start position:0% it was created for. The catch is the 00:02:05.194 --> 00:02:06.475 align:start position:0% it was created for. The catch is the cost. <00:02:06.129>One 00:02:06.475 --> 00:02:06.503 align:start position:0% cost. One 00:02:06.503 --> 00:02:07.598 align:start position:0% cost. One comment <00:02:07.569>stated 00:02:07.598 --> 00:02:07.627 align:start position:0% comment stated 00:02:07.627 --> 00:02:08.491 align:start position:0% comment stated that <00:02:07.732>Fable <00:02:07.837>5.5 <00:02:07.942>would <00:02:08.047>have <00:02:08.152>to <00:02:08.257>be 00:02:08.491 --> 00:02:08.520 align:start position:0% that Fable 5.5 would have to be 00:02:08.520 --> 00:02:09.643 align:start position:0% that Fable 5.5 would have to be truly <00:02:08.995>outstanding <00:02:09.470>to 00:02:09.643 --> 00:02:09.672 align:start position:0% truly outstanding to 00:02:09.672 --> 00:02:10.507 align:start position:0% truly outstanding to justify 00:02:10.507 --> 00:02:10.536 align:start position:0% justify 00:02:10.536 --> 00:02:11.659 align:start position:0% justify using <00:02:10.896>it <00:02:11.256>over 00:02:11.659 --> 00:02:11.688 align:start position:0% using it over 00:02:11.688 --> 00:02:12.466 align:start position:0% using it over Opus, <00:02:12.437>given 00:02:12.466 --> 00:02:12.495 align:start position:0% Opus, given 00:02:12.495 --> 00:02:13.589 align:start position:0% Opus, given how <00:02:12.802>much <00:02:13.109>resources <00:02:13.416>it 00:02:13.589 --> 00:02:13.618 align:start position:0% how much resources it 00:02:13.618 --> 00:02:14.616 align:start position:0% how much resources it consumes. <00:02:13.836>If <00:02:14.054>you <00:02:14.272>are <00:02:14.490>a 00:02:14.616 --> 00:02:14.647 align:start position:0% consumes. If you are a 00:02:14.647 --> 00:02:15.810 align:start position:0% consumes. If you are a regular <00:02:15.244>subscriber 00:02:15.810 --> 00:02:15.841 align:start position:0% regular subscriber 00:02:15.841 --> 00:02:16.815 align:start position:0% regular subscriber without <00:02:16.312>API <00:02:16.783>access, 00:02:16.815 --> 00:02:16.847 align:start position:0% without API access, 00:02:16.847 --> 00:02:17.632 align:start position:0% without API access, a <00:02:17.051>few <00:02:17.255>long 00:02:17.632 --> 00:02:17.663 align:start position:0% a few long 00:02:17.663 --> 00:02:18.512 align:start position:0% a few long requests <00:02:18.135>can 00:02:18.512 --> 00:02:18.543 align:start position:0% requests can 00:02:18.543 --> 00:02:19.486 align:start position:0% requests can exhaust <00:02:18.998>your <00:02:19.453>5- 00:02:19.486 --> 00:02:20.654 align:start position:0% exhaust your 5- hour <00:02:19.944>quota. <00:02:20.402>Therefore, 00:02:20.654 --> 00:02:20.685 align:start position:0% hour quota. Therefore, 00:02:20.685 --> 00:02:21.879 align:start position:0% hour quota. Therefore, at <00:02:20.978>the <00:02:21.271>initial <00:02:21.564>stage, 00:02:21.879 --> 00:02:22.915 align:start position:0% at the initial stage, 00:02:22.915 --> 00:02:22.947 align:start position:0% 00:02:22.947 --> 00:02:23.669 align:start position:0% only 00:02:23.669 --> 00:02:23.700 align:start position:0% only 00:02:23.700 --> 00:02:24.925 align:start position:0% only corporate <00:02:24.485>clients 00:02:24.925 --> 00:02:25.836 align:start position:0% corporate clients willing <00:02:25.160>to <00:02:25.395>pay 00:02:25.836 --> 00:02:25.867 align:start position:0% willing to pay 00:02:25.867 --> 00:02:26.904 align:start position:0% willing to pay more <00:02:26.094>than <00:02:26.321>$50 <00:02:26.548>for <00:02:26.775>a 00:02:26.904 --> 00:02:26.935 align:start position:0% more than $50 for a 00:02:26.935 --> 00:02:28.607 align:start position:0% more than $50 for a million <00:02:27.137>tokens <00:02:27.339>will <00:02:27.541>be <00:02:27.743>able <00:02:27.945>to <00:02:28.147>afford <00:02:28.349>it. <00:02:28.551>Opus, 00:02:28.607 --> 00:02:28.658 align:start position:0% million tokens will be able to afford it. Opus, 00:02:28.658 --> 00:02:30.088 align:start position:0% million tokens will be able to afford it. Opus, Sonnet, <00:02:28.882>and <00:02:29.106>Haiku <00:02:29.330>5.5 <00:02:29.554>are <00:02:29.778>already 00:02:30.088 --> 00:02:30.139 align:start position:0% Sonnet, and Haiku 5.5 are already 00:02:30.139 --> 00:02:31.773 align:start position:0% Sonnet, and Haiku 5.5 are already out <00:02:30.547>and <00:02:30.955>working 00:02:31.773 --> 00:02:31.824 align:start position:0% out and working 00:02:31.824 --> 00:02:33.191 align:start position:0% out and working great. <00:02:32.238>Fable <00:02:32.652>is <00:02:33.066>the 00:02:33.191 --> 00:02:33.222 align:start position:0% great. Fable is the 00:02:33.222 --> 00:02:34.183 align:start position:0% great. Fable is the final <00:02:33.485>piece <00:02:33.748>of 00:02:34.183 --> 00:02:34.214 align:start position:0% final piece of 00:02:34.214 --> 00:02:35.051 align:start position:0% final piece of the <00:02:34.369>puzzle, <00:02:34.524>the <00:02:34.679>model <00:02:34.834>with <00:02:34.989>the 00:02:35.051 --> 00:02:35.082 align:start position:0% the puzzle, the model with the 00:02:35.082 --> 00:02:36.167 align:start position:0% the puzzle, the model with the most <00:02:35.609>unusual 00:02:36.167 --> 00:02:36.198 align:start position:0% most unusual 00:02:36.198 --> 00:02:37.097 align:start position:0% most unusual positioning <00:02:37.035>and 00:02:37.097 --> 00:02:37.128 align:start position:0% positioning and 00:02:37.128 --> 00:02:37.562 align:start position:0% positioning and least 00:02:37.562 --> 00:02:37.593 align:start position:0% least 00:02:37.593 --> 00:02:38.523 align:start position:0% least predictable 00:02:38.523 --> 00:02:38.554 align:start position:0% predictable 00:02:38.554 --> 00:02:39.608 align:start position:0% predictable capabilities <00:02:38.988>in <00:02:39.422>this 00:02:39.608 --> 00:02:39.639 align:start position:0% capabilities in this 00:02:39.639 --> 00:02:40.707 align:start position:0% capabilities in this series. <00:02:40.073>Once <00:02:40.507>it 00:02:40.707 --> 00:02:40.740 align:start position:0% series. Once it 00:02:40.740 --> 00:02:41.673 align:start position:0% series. Once it launches, <00:02:41.373>Anthropic 00:02:41.673 --> 00:02:41.706 align:start position:0% launches, Anthropic 00:02:41.706 --> 00:02:42.472 align:start position:0% launches, Anthropic will <00:02:41.872>cover <00:02:42.038>the <00:02:42.204>entire 00:02:42.472 --> 00:02:42.506 align:start position:0% will cover the entire 00:02:42.506 --> 00:02:43.672 align:start position:0% will cover the entire current <00:02:42.989>market, <00:02:43.472>from 00:02:43.672 --> 00:02:43.705 align:start position:0% current market, from 00:02:43.705 --> 00:02:45.046 align:start position:0% current market, from start <00:02:44.023>to <00:02:44.341>finish. <00:02:44.659>And <00:02:44.977>the 00:02:45.046 --> 00:02:45.080 align:start position:0% start to finish. And the 00:02:45.080 --> 00:02:45.911 align:start position:0% start to finish. And the timing <00:02:45.773>doesn't 00:02:45.911 --> 00:02:45.946 align:start position:0% timing doesn't 00:02:45.946 --> 00:02:47.262 align:start position:0% timing doesn't seem <00:02:47.227>random. 00:02:47.262 --> 00:02:47.431 align:start position:0% seem random. 00:02:47.431 --> 00:02:49.103 align:start position:0% seem random. It <00:02:47.729>looks <00:02:48.027>like <00:02:48.325>an <00:02:48.623>ambush. <00:02:48.921>Argon 00:02:49.103 --> 00:02:49.139 align:start position:0% It looks like an ambush. Argon 00:02:49.139 --> 00:02:50.426 align:start position:0% It looks like an ambush. Argon gets <00:02:49.425>its <00:02:49.711>moment <00:02:49.997>of 00:02:50.426 --> 00:02:50.462 align:start position:0% gets its moment of 00:02:50.462 --> 00:02:51.535 align:start position:0% gets its moment of glory, <00:02:50.748>and <00:02:51.034>before <00:02:51.320>the 00:02:51.535 --> 00:02:51.570 align:start position:0% glory, and before the 00:02:51.570 --> 00:02:52.893 align:start position:0% glory, and before the hype <00:02:51.939>dies <00:02:52.308>down, <00:02:52.677>Fable 00:02:52.893 --> 00:02:52.929 align:start position:0% hype dies down, Fable 00:02:52.929 --> 00:02:54.323 align:start position:0% hype dies down, Fable is <00:02:53.064>already <00:02:53.199>preparing <00:02:53.334>to <00:02:53.469>steal <00:02:53.604>the 00:02:54.323 --> 00:02:54.359 align:start position:0% is already preparing to steal the 00:02:54.359 --> 00:02:55.585 align:start position:0% is already preparing to steal the spotlight. <00:02:55.278>One 00:02:55.585 --> 00:02:55.616 align:start position:0% spotlight. One 00:02:55.616 --> 00:02:56.780 align:start position:0% spotlight. One thought <00:02:55.884>leader <00:02:56.152>said: <00:02:56.420>"Anthropic <00:02:56.688>is 00:02:56.780 --> 00:02:57.669 align:start position:0% thought leader said: "Anthropic is the <00:02:57.209>most <00:02:57.638>ruthless, 00:02:57.669 --> 00:02:57.700 align:start position:0% the most ruthless, 00:02:57.700 --> 00:02:58.374 align:start position:0% the most ruthless, strategically 00:02:58.374 --> 00:02:58.405 align:start position:0% strategically 00:02:58.405 --> 00:02:59.140 align:start position:0% strategically calculated 00:02:59.140 --> 00:02:59.171 align:start position:0% calculated 00:02:59.171 --> 00:03:00.336 align:start position:0% calculated company <00:02:59.454>in <00:02:59.737>the <00:03:00.020>AI ​​<00:03:00.303>industry, 00:03:00.336 --> 00:03:00.366 align:start position:0% company in the AI ​​industry, 00:03:00.366 --> 00:03:01.133 align:start position:0% company in the AI ​​industry, they <00:03:00.504>are <00:03:00.642>constantly 00:03:01.133 --> 00:03:01.163 align:start position:0% they are constantly 00:03:01.163 --> 00:03:01.838 align:start position:0% they are constantly crushing 00:03:01.838 --> 00:03:01.868 align:start position:0% crushing 00:03:01.868 --> 00:03:02.635 align:start position:0% crushing the <00:03:02.220>competition <00:03:02.572>and 00:03:02.635 --> 00:03:02.665 align:start position:0% the competition and 00:03:02.665 --> 00:03:03.769 align:start position:0% the competition and taking <00:03:02.930>all <00:03:03.195>the <00:03:03.460>attention 00:03:03.769 --> 00:03:03.799 align:start position:0% taking all the attention 00:03:03.799 --> 00:03:04.780 align:start position:0% taking all the attention , 00:03:04.780 --> 00:03:04.811 align:start position:0% , 00:03:04.811 --> 00:03:05.761 align:start position:0% , they <00:03:04.949>seem <00:03:05.087>determined 00:03:05.761 --> 00:03:05.792 align:start position:0% they seem determined 00:03:05.792 --> 00:03:06.864 align:start position:0% they seem determined to <00:03:06.083>stay <00:03:06.374>ahead 00:03:06.864 --> 00:03:06.895 align:start position:0% to stay ahead 00:03:06.895 --> 00:03:07.907 align:start position:0% to stay ahead until <00:03:07.222>they <00:03:07.549>go <00:03:07.876>public." 00:03:07.907 --> 00:03:08.007 align:start position:0% until they go public." 00:03:08.007 --> 00:03:08.709 align:start position:0% until they go public." Industry <00:03:08.397>experts 00:03:08.709 --> 00:03:08.735 align:start position:0% Industry experts 00:03:08.735 --> 00:03:09.620 align:start position:0% Industry experts believe <00:03:08.943>that <00:03:09.151>this <00:03:09.359>is <00:03:09.567>the 00:03:09.620 --> 00:03:09.646 align:start position:0% believe that this is the 00:03:09.646 --> 00:03:10.478 align:start position:0% believe that this is the point: <00:03:09.958>to <00:03:10.270>have 00:03:10.478 --> 00:03:10.504 align:start position:0% point: to have 00:03:10.504 --> 00:03:11.233 align:start position:0% point: to have more <00:03:10.790>time <00:03:11.076>to 00:03:11.233 --> 00:03:11.259 align:start position:0% more time to 00:03:11.259 --> 00:03:12.117 align:start position:0% more time to build <00:03:11.857>momentum 00:03:12.117 --> 00:03:12.143 align:start position:0% build momentum 00:03:12.143 --> 00:03:12.898 align:start position:0% build momentum before <00:03:12.286>the <00:03:12.429>public 00:03:12.898 --> 00:03:12.924 align:start position:0% before the public 00:03:12.924 --> 00:03:13.809 align:start position:0% before the public offering <00:03:13.353>of <00:03:13.782>shares. 00:03:13.809 --> 00:03:13.909 align:start position:0% offering of shares. 00:03:13.909 --> 00:03:15.034 align:start position:0% offering of shares. OpenAI <00:03:14.081>is <00:03:14.253>also <00:03:14.425>not <00:03:14.597>sitting 00:03:15.034 --> 00:03:15.070 align:start position:0% OpenAI is also not sitting 00:03:15.070 --> 00:03:16.185 align:start position:0% OpenAI is also not sitting idly <00:03:15.556>by. <00:03:16.042>According 00:03:16.185 --> 00:03:16.221 align:start position:0% idly by. According 00:03:16.221 --> 00:03:17.408 align:start position:0% idly by. According to <00:03:16.761>an 00:03:17.408 --> 00:03:17.444 align:start position:0% to an 00:03:17.444 --> 00:03:18.703 align:start position:0% to an industry <00:03:17.687>insider <00:03:17.930>at <00:03:18.173>Synthwave, <00:03:18.416>they 00:03:18.703 --> 00:03:18.739 align:start position:0% industry insider at Synthwave, they 00:03:18.739 --> 00:03:19.854 align:start position:0% industry insider at Synthwave, they plan <00:03:19.045>to <00:03:19.351>heavily 00:03:19.854 --> 00:03:19.890 align:start position:0% plan to heavily 00:03:19.890 --> 00:03:21.149 align:start position:0% plan to heavily promote <00:03:20.411>6.1 <00:03:20.932>Astra 00:03:21.149 --> 00:03:21.185 align:start position:0% promote 6.1 Astra 00:03:21.185 --> 00:03:22.444 align:start position:0% promote 6.1 Astra during <00:03:21.563>their <00:03:21.941>28-day 00:03:22.444 --> 00:03:22.480 align:start position:0% during their 28-day 00:03:22.480 --> 00:03:23.595 align:start position:0% during their 28-day release <00:03:23.092>marathon 00:03:23.595 --> 00:03:23.631 align:start position:0% release marathon 00:03:23.631 --> 00:03:24.944 align:start position:0% release marathon this <00:03:23.890>month. <00:03:24.149>Although <00:03:24.408>there <00:03:24.667>are 00:03:24.944 --> 00:03:24.979 align:start position:0% this month. Although there are 00:03:24.979 --> 00:03:26.188 align:start position:0% this month. Although there are rumors <00:03:25.324>that <00:03:25.669>the <00:03:26.014>Astra 00:03:26.188 --> 00:03:26.223 align:start position:0% rumors that the Astra 00:03:26.223 --> 00:03:27.225 align:start position:0% rumors that the Astra was <00:03:26.534>delayed 00:03:27.225 --> 00:03:27.260 align:start position:0% was delayed 00:03:27.260 --> 00:03:28.573 align:start position:0% was delayed earlier <00:03:27.536>after <00:03:27.812>failing <00:03:28.088>a 00:03:28.573 --> 00:03:28.608 align:start position:0% earlier after failing a 00:03:28.608 --> 00:03:29.946 align:start position:0% earlier after failing a safety <00:03:28.912>assessment. <00:03:29.216>Whatever <00:03:29.520>the <00:03:29.824>case 00:03:29.946 --> 00:03:29.976 align:start position:0% safety assessment. Whatever the case 00:03:29.976 --> 00:03:30.607 align:start position:0% safety assessment. Whatever the case , 00:03:30.607 --> 00:03:30.637 align:start position:0% , 00:03:30.637 --> 00:03:31.810 align:start position:0% , passions <00:03:30.847>are <00:03:31.057>running <00:03:31.267>high 00:03:31.810 --> 00:03:32.952 align:start position:0% passions are running high , <00:03:32.032>and <00:03:32.254>people <00:03:32.476>are <00:03:32.698>already <00:03:32.920>joking: 00:03:32.952 --> 00:03:32.982 align:start position:0% , and people are already joking: 00:03:32.982 --> 00:03:34.064 align:start position:0% , and people are already joking: "Enjoy <00:03:33.764>your 00:03:34.064 --> 00:03:34.094 align:start position:0% "Enjoy your 00:03:34.094 --> 00:03:34.996 align:start position:0% "Enjoy your short <00:03:34.374>time <00:03:34.654>in <00:03:34.934>the 00:03:34.996 --> 00:03:35.026 align:start position:0% short time in the 00:03:35.026 --> 00:03:35.748 align:start position:0% short time in the spotlight, 00:03:35.748 --> 00:03:35.778 align:start position:0% spotlight, 00:03:35.778 --> 00:03:37.196 align:start position:0% spotlight, little <00:03:36.106>Argon." <00:03:36.434>A <00:03:36.762>few 00:03:37.196 --> 00:03:37.232 align:start position:0% little Argon." A few 00:03:37.232 --> 00:03:38.318 align:start position:0% little Argon." A few days <00:03:37.558>ago, <00:03:37.884>observant 00:03:38.318 --> 00:03:38.354 align:start position:0% days ago, observant 00:03:38.354 --> 00:03:39.766 align:start position:0% days ago, observant users <00:03:39.187>noticed 00:03:39.766 --> 00:03:40.598 align:start position:0% users noticed that <00:03:40.031>Claude <00:03:40.296>Opus <00:03:40.561>5.5 00:03:40.598 --> 00:03:40.635 align:start position:0% that Claude Opus 5.5 00:03:40.635 --> 00:03:41.793 align:start position:0% that Claude Opus 5.5 seemed <00:03:40.876>to <00:03:41.117>be <00:03:41.358>secretly 00:03:41.793 --> 00:03:41.829 align:start position:0% seemed to be secretly 00:03:41.829 --> 00:03:42.770 align:start position:0% seemed to be secretly redirecting 00:03:42.770 --> 00:03:42.806 align:start position:0% redirecting 00:03:42.806 --> 00:03:44.182 align:start position:0% redirecting super-complex <00:03:43.421>queries <00:03:44.036>to 00:03:44.182 --> 00:03:44.218 align:start position:0% super-complex queries to 00:03:44.218 --> 00:03:45.449 align:start position:0% super-complex queries to some <00:03:44.815>mysterious, <00:03:45.412>more 00:03:45.449 --> 00:03:45.485 align:start position:0% some mysterious, more 00:03:45.485 --> 00:03:46.928 align:start position:0% some mysterious, more powerful <00:03:46.152>model. <00:03:46.819>This 00:03:46.928 --> 00:03:46.955 align:start position:0% powerful model. This 00:03:46.955 --> 00:03:47.583 align:start position:0% powerful model. This coincides <00:03:47.528>with 00:03:47.583 --> 00:03:47.610 align:start position:0% coincides with 00:03:47.610 --> 00:03:48.566 align:start position:0% coincides with reports 00:03:48.566 --> 00:03:48.593 align:start position:0% reports 00:03:48.593 --> 00:03:49.331 align:start position:0% reports that <00:03:48.880>Fable <00:03:49.167>5.5 00:03:49.331 --> 00:03:49.358 align:start position:0% that Fable 5.5 00:03:49.358 --> 00:03:50.396 align:start position:0% that Fable 5.5 is <00:03:49.558>already <00:03:49.758>in <00:03:49.958>early 00:03:50.396 --> 00:03:50.423 align:start position:0% is already in early 00:03:50.423 --> 00:03:51.488 align:start position:0% is already in early access, <00:03:50.687>a <00:03:50.951>"gray" <00:03:51.215>release 00:03:51.488 --> 00:03:51.515 align:start position:0% access, a "gray" release 00:03:51.515 --> 00:03:52.498 align:start position:0% access, a "gray" release for <00:03:51.788>some <00:03:52.061>companies 00:03:52.498 --> 00:03:52.526 align:start position:0% for some companies 00:03:52.526 --> 00:03:53.563 align:start position:0% for some companies and <00:03:52.771>individuals, <00:03:53.016>and 00:03:53.563 --> 00:03:53.591 align:start position:0% and individuals, and 00:03:53.591 --> 00:03:54.273 align:start position:0% and individuals, and only <00:03:53.836>final 00:03:54.273 --> 00:03:54.301 align:start position:0% only final 00:03:54.301 --> 00:03:55.147 align:start position:0% only final testing <00:03:54.492>remains <00:03:54.683>before <00:03:54.874>the 00:03:55.147 --> 00:03:55.174 align:start position:0% testing remains before the 00:03:55.174 --> 00:03:56.158 align:start position:0% testing remains before the public <00:03:56.130>launch. 00:03:56.158 --> 00:03:56.258 align:start position:0% public launch. 00:03:56.258 --> 00:03:57.263 align:start position:0% public launch. People <00:03:56.355>have <00:03:56.452>even <00:03:56.549>come <00:03:56.646>up <00:03:56.743>with <00:03:56.840>a 00:03:57.263 --> 00:03:57.289 align:start position:0% People have even come up with a 00:03:57.289 --> 00:03:58.321 align:start position:0% People have even come up with a way <00:03:57.597>to <00:03:57.905>check <00:03:58.213>if 00:03:58.321 --> 00:03:58.348 align:start position:0% way to check if 00:03:58.348 --> 00:03:59.450 align:start position:0% way to check if you're <00:03:58.531>on <00:03:58.714>topic. <00:03:58.897>Turn 00:03:59.450 --> 00:03:59.481 align:start position:0% you're on topic. Turn 00:03:59.481 --> 00:04:00.280 align:start position:0% you're on topic. Turn Opus <00:03:59.624>5.5 <00:03:59.767>on <00:03:59.910>high 00:04:00.280 --> 00:04:00.310 align:start position:0% Opus 5.5 on high 00:04:00.310 --> 00:04:01.508 align:start position:0% Opus 5.5 on high power <00:04:00.893>and <00:04:01.476>type: 00:04:01.508 --> 00:04:01.539 align:start position:0% power and type: 00:04:01.539 --> 00:04:02.645 align:start position:0% power and type: "Do <00:04:01.825>you <00:04:02.111>know <00:04:02.397>what 00:04:02.645 --> 00:04:02.675 align:start position:0% "Do you know what 00:04:02.675 --> 00:04:03.289 align:start position:0% "Do you know what Voxel <00:04:02.841>Bench <00:04:03.007>is?" <00:04:03.173>Don't 00:04:03.289 --> 00:04:03.317 align:start position:0% Voxel Bench is?" Don't 00:04:03.317 --> 00:04:04.347 align:start position:0% Voxel Bench is?" Don't use <00:04:04.318>search. 00:04:04.347 --> 00:04:04.447 align:start position:0% use search. 00:04:04.447 --> 00:04:06.076 align:start position:0% use search. Then: <00:04:04.974>"Do <00:04:05.501>you <00:04:06.028>know 00:04:06.076 --> 00:04:06.124 align:start position:0% Then: "Do you know 00:04:06.124 --> 00:04:07.782 align:start position:0% Then: "Do you know who <00:04:06.636>@legit_api <00:04:07.148>is?" <00:04:07.660>Don't 00:04:07.782 --> 00:04:07.812 align:start position:0% who @legit_api is?" Don't 00:04:07.812 --> 00:04:08.893 align:start position:0% who @legit_api is?" Don't use <00:04:08.862>search. 00:04:08.893 --> 00:04:08.993 align:start position:0% use search. 00:04:08.993 --> 00:04:10.563 align:start position:0% use search. Opus <00:04:09.300>5.5 <00:04:09.607>has <00:04:09.914>an <00:04:10.221>earlier 00:04:10.563 --> 00:04:10.605 align:start position:0% Opus 5.5 has an earlier 00:04:10.605 --> 00:04:12.232 align:start position:0% Opus 5.5 has an earlier knowledge <00:04:11.091>cutoff <00:04:11.577>date. <00:04:12.063>So 00:04:12.232 --> 00:04:12.260 align:start position:0% knowledge cutoff date. So 00:04:12.260 --> 00:04:13.301 align:start position:0% knowledge cutoff date. So usually <00:04:12.668>he <00:04:13.076>can't 00:04:13.301 --> 00:04:13.329 align:start position:0% usually he can't 00:04:13.329 --> 00:04:14.426 align:start position:0% usually he can't answer <00:04:13.849>any <00:04:14.369>of 00:04:14.426 --> 00:04:14.454 align:start position:0% answer any of 00:04:14.454 --> 00:04:15.561 align:start position:0% answer any of them. <00:04:14.675>But <00:04:14.896>if <00:04:15.117>you <00:04:15.338>are 00:04:15.561 --> 00:04:15.598 align:start position:0% them. But if you are 00:04:15.598 --> 00:04:16.741 align:start position:0% them. But if you are redirected <00:04:16.095>to <00:04:16.592>a 00:04:16.741 --> 00:04:16.778 align:start position:0% redirected to a 00:04:16.778 --> 00:04:17.847 align:start position:0% redirected to a new <00:04:17.201>model, <00:04:17.624>he 00:04:17.847 --> 00:04:17.883 align:start position:0% new model, he 00:04:17.883 --> 00:04:19.284 align:start position:0% new model, he will <00:04:18.126>tell <00:04:18.369>you <00:04:18.612>that <00:04:18.855>Voxel <00:04:19.098>Bench 00:04:19.284 --> 00:04:20.169 align:start position:0% will tell you that Voxel Bench is <00:04:19.579>a 00:04:20.169 --> 00:04:20.205 align:start position:0% is a 00:04:20.205 --> 00:04:21.090 align:start position:0% is a spatial 00:04:21.090 --> 00:04:21.127 align:start position:0% spatial 00:04:21.127 --> 00:04:22.085 align:start position:0% spatial thinking <00:04:21.409>benchmark <00:04:21.691>where <00:04:21.973>LLMs 00:04:22.085 --> 00:04:22.122 align:start position:0% thinking benchmark where LLMs 00:04:22.122 --> 00:04:23.559 align:start position:0% thinking benchmark where LLMs generate <00:04:22.822>3D 00:04:23.559 --> 00:04:23.596 align:start position:0% generate 3D 00:04:23.596 --> 00:04:24.776 align:start position:0% generate 3D voxels, <00:04:23.863>and <00:04:24.130>@legit_api <00:04:24.397>is <00:04:24.664>a 00:04:24.776 --> 00:04:25.587 align:start position:0% voxels, and @legit_api is a blogger <00:04:25.292>who 00:04:25.587 --> 00:04:25.623 align:start position:0% blogger who 00:04:25.623 --> 00:04:26.913 align:start position:0% blogger who specializes <00:04:26.766>in 00:04:26.913 --> 00:04:26.950 align:start position:0% specializes in 00:04:26.950 --> 00:04:28.093 align:start position:0% specializes in news <00:04:27.115>about <00:04:27.280>models <00:04:27.445>that <00:04:27.610>have <00:04:27.775>not <00:04:27.940>yet 00:04:28.093 --> 00:04:28.130 align:start position:0% news about models that have not yet 00:04:28.130 --> 00:04:29.986 align:start position:0% news about models that have not yet been <00:04:28.889>released. <00:04:29.648>So 00:04:29.986 --> 00:04:30.961 align:start position:0% been released. So what <00:04:30.297>can <00:04:30.608>this <00:04:30.919>thing, 00:04:30.961 --> 00:04:31.003 align:start position:0% what can this thing, 00:04:31.003 --> 00:04:32.571 align:start position:0% what can this thing, most <00:04:31.511>likely <00:04:32.019>Fable <00:04:32.527>5.5, 00:04:32.571 --> 00:04:32.613 align:start position:0% most likely Fable 5.5, 00:04:32.613 --> 00:04:33.757 align:start position:0% most likely Fable 5.5, actually 00:04:33.757 --> 00:04:33.800 align:start position:0% actually 00:04:33.800 --> 00:04:35.880 align:start position:0% actually do? <00:04:34.630>One <00:04:35.460>person 00:04:35.880 --> 00:04:35.915 align:start position:0% do? One person 00:04:35.915 --> 00:04:36.926 align:start position:0% do? One person said <00:04:36.281>that <00:04:36.647>she 00:04:36.926 --> 00:04:36.961 align:start position:0% said that she 00:04:36.961 --> 00:04:38.251 align:start position:0% said that she created <00:04:37.257>a <00:04:37.553>full-fledged 00:04:38.251 --> 00:04:38.286 align:start position:0% created a full-fledged 00:04:38.286 --> 00:04:39.612 align:start position:0% created a full-fledged MMORPG <00:04:38.608>game <00:04:38.930>in <00:04:39.252>5 <00:04:39.574>hours. 00:04:39.612 --> 00:04:40.389 align:start position:0% MMORPG game in 5 hours. 00:04:40.389 --> 00:04:41.621 align:start position:0% MMORPG game in 5 hours. Maps, <00:04:41.581>characters, 00:04:41.621 --> 00:04:41.661 align:start position:0% Maps, characters, 00:04:41.661 --> 00:04:43.171 align:start position:0% Maps, characters, interactions—all <00:04:42.774>through 00:04:43.171 --> 00:04:43.211 align:start position:0% interactions—all through 00:04:43.211 --> 00:04:45.305 align:start position:0% interactions—all through code. <00:04:43.691>Another <00:04:44.171>asked <00:04:44.651>for 00:04:45.305 --> 00:04:45.345 align:start position:0% code. Another asked for 00:04:45.345 --> 00:04:46.202 align:start position:0% code. Another asked for an <00:04:45.549>explanation <00:04:45.753>of <00:04:45.957>OpenAI's <00:04:46.161>199- 00:04:46.202 --> 00:04:47.792 align:start position:0% an explanation of OpenAI's 199- page <00:04:46.487>proof <00:04:46.772>of <00:04:47.057>the 00:04:47.792 --> 00:04:47.833 align:start position:0% page proof of the 00:04:47.833 --> 00:04:48.934 align:start position:0% page proof of the quasi- 00:04:48.934 --> 00:04:48.975 align:start position:0% quasi- 00:04:48.975 --> 00:04:50.740 align:start position:0% quasi- Riemann <00:04:49.826>hypothesis. <00:04:50.677>And 00:04:50.740 --> 00:04:50.771 align:start position:0% Riemann hypothesis. And 00:04:50.771 --> 00:04:51.802 align:start position:0% Riemann hypothesis. And instead <00:04:51.005>of <00:04:51.239>explaining, 00:04:51.802 --> 00:04:51.833 align:start position:0% instead of explaining, 00:04:51.833 --> 00:04:52.989 align:start position:0% instead of explaining, the <00:04:52.208>model <00:04:52.583>wrote <00:04:52.958>code 00:04:52.989 --> 00:04:53.020 align:start position:0% the model wrote code 00:04:53.020 --> 00:04:53.988 align:start position:0% the model wrote code that <00:04:53.160>rendered <00:04:53.300>a 00:04:53.988 --> 00:04:54.020 align:start position:0% that rendered a 00:04:54.020 --> 00:04:55.238 align:start position:0% that rendered a three-dimensional <00:04:54.675>animated 00:04:55.238 --> 00:04:55.269 align:start position:0% three-dimensional animated 00:04:55.269 --> 00:04:56.237 align:start position:0% three-dimensional animated video <00:04:55.612>lasting 00:04:56.237 --> 00:04:56.268 align:start position:0% video lasting 00:04:56.268 --> 00:04:57.666 align:start position:0% video lasting several <00:04:56.782>minutes. <00:04:57.296>Someone 00:04:57.666 --> 00:04:57.702 align:start position:0% several minutes. Someone 00:04:57.702 --> 00:04:59.142 align:start position:0% several minutes. Someone even <00:04:58.052>threw <00:04:58.402>her <00:04:58.752>a <00:04:59.102>160- 00:04:59.142 --> 00:05:00.581 align:start position:0% even threw her a 160- page <00:04:59.400>proof <00:04:59.658>of <00:04:59.916>the 00:05:00.581 --> 00:05:00.618 align:start position:0% page proof of the 00:05:00.618 --> 00:05:02.021 align:start position:0% page proof of the Navier-Stokes <00:05:01.984>equations. 00:05:02.021 --> 00:05:02.437 align:start position:0% Navier-Stokes equations. 00:05:02.437 --> 00:05:04.473 align:start position:0% Navier-Stokes equations. But <00:05:02.927>the <00:05:03.417>real <00:05:03.907>sensation 00:05:04.473 --> 00:05:04.529 align:start position:0% But the real sensation 00:05:04.529 --> 00:05:06.339 align:start position:0% But the real sensation was <00:05:04.823>caused <00:05:05.117>by <00:05:05.411>a <00:05:05.705>post <00:05:05.999>from 00:05:06.339 --> 00:05:06.396 align:start position:0% was caused by a post from 00:05:06.396 --> 00:05:08.685 align:start position:0% was caused by a post from @nft_chin. <00:05:06.771>In <00:05:07.146>Claude <00:05:07.521>code, <00:05:07.896>the <00:05:08.271>model 00:05:08.685 --> 00:05:08.720 align:start position:0% @nft_chin. In Claude code, the model 00:05:08.720 --> 00:05:09.650 align:start position:0% @nft_chin. In Claude code, the model effectively <00:05:09.305>became 00:05:09.650 --> 00:05:09.684 align:start position:0% effectively became 00:05:09.684 --> 00:05:10.407 align:start position:0% effectively became the <00:05:10.011>director <00:05:10.338>and 00:05:10.407 --> 00:05:10.442 align:start position:0% the director and 00:05:10.442 --> 00:05:11.234 align:start position:0% the director and spent <00:05:10.803>2 <00:05:11.164>and 00:05:11.234 --> 00:05:11.268 align:start position:0% spent 2 and 00:05:11.268 --> 00:05:12.508 align:start position:0% spent 2 and a <00:05:11.819>half <00:05:12.370>hours 00:05:12.508 --> 00:05:12.542 align:start position:0% a half hours 00:05:12.542 --> 00:05:13.817 align:start position:0% a half hours writing <00:05:12.869>a <00:05:13.196>poetic 00:05:13.817 --> 00:05:13.851 align:start position:0% writing a poetic 00:05:13.851 --> 00:05:14.884 align:start position:0% writing a poetic animation <00:05:14.023>in <00:05:14.195>the <00:05:14.367>style <00:05:14.539>of 00:05:14.884 --> 00:05:14.919 align:start position:0% animation in the style of 00:05:14.919 --> 00:05:16.124 align:start position:0% animation in the style of Wong <00:05:15.349>Kar-wai, <00:05:15.779>almost 00:05:16.124 --> 00:05:16.158 align:start position:0% Wong Kar-wai, almost 00:05:16.158 --> 00:05:18.068 align:start position:0% Wong Kar-wai, almost entirely <00:05:16.706>in <00:05:17.254>code. <00:05:17.802>She 00:05:18.068 --> 00:05:18.101 align:start position:0% entirely in code. She 00:05:18.101 --> 00:05:19.157 align:start position:0% entirely in code. She created <00:05:18.409>an <00:05:18.717>intro <00:05:19.025>with 00:05:19.157 --> 00:05:20.280 align:start position:0% created an intro with 00:05:20.280 --> 00:05:20.313 align:start position:0% 00:05:20.313 --> 00:05:21.535 align:start position:0% hand-drawn <00:05:20.907>brush <00:05:21.501>strokes, 00:05:21.535 --> 00:05:21.568 align:start position:0% hand-drawn brush strokes, 00:05:21.568 --> 00:05:22.228 align:start position:0% hand-drawn brush strokes, trilingual 00:05:22.228 --> 00:05:22.261 align:start position:0% trilingual 00:05:22.261 --> 00:05:22.922 align:start position:0% trilingual subtitles 00:05:22.922 --> 00:05:22.955 align:start position:0% subtitles 00:05:22.955 --> 00:05:23.648 align:start position:0% subtitles in <00:05:23.615>Chinese, 00:05:23.648 --> 00:05:23.681 align:start position:0% in Chinese, 00:05:23.681 --> 00:05:24.573 align:start position:0% in Chinese, English, <00:05:24.441>and 00:05:24.573 --> 00:05:24.606 align:start position:0% English, and 00:05:24.606 --> 00:05:25.596 align:start position:0% English, and Japanese, <00:05:24.936>a <00:05:25.266>pure 00:05:25.596 --> 00:05:25.629 align:start position:0% Japanese, a pure 00:05:25.629 --> 00:05:26.620 align:start position:0% Japanese, a pure red <00:05:26.091>background, <00:05:26.553>and 00:05:26.620 --> 00:05:26.653 align:start position:0% red background, and 00:05:26.653 --> 00:05:27.875 align:start position:0% red background, and white <00:05:27.247>text: <00:05:27.841>"So, 00:05:27.875 --> 00:05:27.908 align:start position:0% white text: "So, 00:05:27.908 --> 00:05:28.766 align:start position:0% white text: "So, I <00:05:28.205>broke <00:05:28.502>my 00:05:28.766 --> 00:05:28.799 align:start position:0% I broke my 00:05:28.799 --> 00:05:29.857 align:start position:0% I broke my promise." 00:05:29.857 --> 00:05:29.887 align:start position:0% promise." 00:05:29.887 --> 00:05:30.596 align:start position:0% promise." Distance <00:05:30.180>markers <00:05:30.473>on 00:05:30.596 --> 00:05:30.627 align:start position:0% Distance markers on 00:05:30.627 --> 00:05:31.367 align:start position:0% Distance markers on isometric 00:05:31.367 --> 00:05:31.398 align:start position:0% isometric 00:05:31.398 --> 00:05:32.045 align:start position:0% isometric platforms, 00:05:32.045 --> 00:05:32.076 align:start position:0% platforms, 00:05:32.076 --> 00:05:32.754 align:start position:0% platforms, storyboards of 00:05:32.754 --> 00:05:32.784 align:start position:0% storyboards of 00:05:32.784 --> 00:05:33.986 align:start position:0% storyboards of colorful <00:05:33.432>striped 00:05:33.986 --> 00:05:34.017 align:start position:0% colorful striped 00:05:34.017 --> 00:05:34.695 align:start position:0% colorful striped buildings 00:05:34.695 --> 00:05:34.726 align:start position:0% buildings 00:05:34.726 --> 00:05:35.404 align:start position:0% buildings crossing 00:05:35.404 --> 00:05:35.435 align:start position:0% crossing 00:05:35.435 --> 00:05:36.513 align:start position:0% crossing railroad <00:05:35.943>tracks, <00:05:36.451>and 00:05:36.513 --> 00:05:37.592 align:start position:0% railroad tracks, and 00:05:37.592 --> 00:05:37.623 align:start position:0% 00:05:37.623 --> 00:05:38.578 align:start position:0% paper 00:05:38.578 --> 00:05:38.609 align:start position:0% paper 00:05:38.609 --> 00:05:40.315 align:start position:0% paper collage <00:05:38.833>style <00:05:39.057>character <00:05:39.281>files. <00:05:39.505>All <00:05:39.729>of <00:05:39.953>this <00:05:40.177>is 00:05:40.315 --> 00:05:40.349 align:start position:0% collage style character files. All of this is 00:05:40.349 --> 00:05:41.200 align:start position:0% collage style character files. All of this is built <00:05:41.064>on 00:05:41.200 --> 00:05:41.234 align:start position:0% built on 00:05:41.234 --> 00:05:41.847 align:start position:0% built on the <00:05:41.427>images <00:05:41.620>of <00:05:41.813>"5 00:05:41.847 --> 00:05:41.881 align:start position:0% the images of "5 00:05:41.881 --> 00:05:42.800 align:start position:0% the images of "5 Centimeters <00:05:42.663>Per 00:05:42.800 --> 00:05:42.834 align:start position:0% Centimeters Per 00:05:42.834 --> 00:05:43.923 align:start position:0% Centimeters Per Second," <00:05:43.009>and <00:05:43.184>the <00:05:43.359>AI ​​<00:05:43.534>made <00:05:43.709>the <00:05:43.884>style, 00:05:43.923 --> 00:05:43.957 align:start position:0% Second," and the AI ​​made the style, 00:05:43.957 --> 00:05:45.216 align:start position:0% Second," and the AI ​​made the style, rhythm, <00:05:44.518>storyboards, <00:05:45.079>and 00:05:45.216 --> 00:05:45.250 align:start position:0% rhythm, storyboards, and 00:05:45.250 --> 00:05:46.101 align:start position:0% rhythm, storyboards, and every <00:05:45.625>artistic 00:05:46.101 --> 00:05:46.135 align:start position:0% every artistic 00:05:46.135 --> 00:05:47.429 align:start position:0% every artistic decision 00:05:47.429 --> 00:05:47.463 align:start position:0% decision 00:05:47.463 --> 00:05:49.554 align:start position:0% decision independently. <00:05:48.252>The <00:05:49.041>reaction 00:05:49.554 --> 00:05:49.591 align:start position:0% independently. The reaction 00:05:49.591 --> 00:05:51.021 align:start position:0% independently. The reaction was <00:05:49.939>something <00:05:50.287>like <00:05:50.635>this: <00:05:50.983>are 00:05:51.021 --> 00:05:51.058 align:start position:0% was something like this: are 00:05:51.058 --> 00:05:52.341 align:start position:0% was something like this: are n't <00:05:52.305>LLMs 00:05:52.341 --> 00:05:52.378 align:start position:0% n't LLMs 00:05:52.378 --> 00:05:53.772 align:start position:0% n't LLMs writing <00:05:52.830>clean <00:05:53.282>code <00:05:53.734>going 00:05:53.772 --> 00:05:53.808 align:start position:0% writing clean code going 00:05:53.808 --> 00:05:55.165 align:start position:0% writing clean code going to <00:05:53.991>destroy <00:05:54.174>video <00:05:54.357>models 00:05:55.165 --> 00:05:55.202 align:start position:0% to destroy video models 00:05:55.202 --> 00:05:56.072 align:start position:0% to destroy video models like <00:05:55.443>Sora? <00:05:55.684>And <00:05:55.925>this 00:05:56.072 --> 00:05:56.109 align:start position:0% like Sora? And this 00:05:56.109 --> 00:05:57.498 align:start position:0% like Sora? And this continues. <00:05:57.198>Someone 00:05:57.498 --> 00:05:57.528 align:start position:0% continues. Someone 00:05:57.528 --> 00:05:58.641 align:start position:0% continues. Someone created <00:05:58.024>an <00:05:58.520>animation 00:05:58.641 --> 00:05:58.671 align:start position:0% created an animation 00:05:58.671 --> 00:05:59.633 align:start position:0% created an animation spanning <00:05:58.891>40,000 <00:05:59.111>years <00:05:59.331>of 00:05:59.633 --> 00:05:59.663 align:start position:0% spanning 40,000 years of 00:05:59.663 --> 00:06:00.656 align:start position:0% spanning 40,000 years of human 00:06:00.656 --> 00:06:00.686 align:start position:0% human 00:06:00.686 --> 00:06:01.858 align:start position:0% human art <00:06:00.858>history, <00:06:01.030>and <00:06:01.202>the <00:06:01.374>conclusion 00:06:01.858 --> 00:06:01.888 align:start position:0% art history, and the conclusion 00:06:01.888 --> 00:06:02.821 align:start position:0% art history, and the conclusion was <00:06:02.078>this: <00:06:02.268>even <00:06:02.458>a 00:06:02.821 --> 00:06:02.851 align:start position:0% was this: even a 00:06:02.851 --> 00:06:03.663 align:start position:0% was this: even a top-notch <00:06:03.046>team <00:06:03.241>of 00:06:03.663 --> 00:06:03.693 align:start position:0% top-notch team of 00:06:03.693 --> 00:06:04.444 align:start position:0% top-notch team of animators <00:06:04.008>couldn't <00:06:04.323>have 00:06:04.444 --> 00:06:04.474 align:start position:0% animators couldn't have 00:06:04.474 --> 00:06:05.467 align:start position:0% animators couldn't have done <00:06:04.764>it <00:06:05.054>in <00:06:05.344>a 00:06:05.467 --> 00:06:05.497 align:start position:0% done it in a 00:06:05.497 --> 00:06:06.806 align:start position:0% done it in a month <00:06:05.840>or <00:06:06.183>two. <00:06:06.526>Another 00:06:06.806 --> 00:06:06.841 align:start position:0% month or two. Another 00:06:06.841 --> 00:06:08.025 align:start position:0% month or two. Another work <00:06:07.328>shows <00:06:07.815>a 00:06:08.025 --> 00:06:08.059 align:start position:0% work shows a 00:06:08.059 --> 00:06:08.895 align:start position:0% work shows a person's <00:06:08.459>entire <00:06:08.859>life, 00:06:08.895 --> 00:06:08.930 align:start position:0% person's entire life, 00:06:08.930 --> 00:06:10.183 align:start position:0% person's entire life, drawn <00:06:09.487>frame <00:06:10.044>by 00:06:10.183 --> 00:06:10.218 align:start position:0% drawn frame by 00:06:10.218 --> 00:06:11.228 align:start position:0% drawn frame by frame <00:06:10.705>in <00:06:11.192>JavaScript. 00:06:11.228 --> 00:06:11.328 align:start position:0% frame in JavaScript. 00:06:11.328 --> 00:06:12.439 align:start position:0% frame in JavaScript. Someone <00:06:11.760>asked <00:06:12.192>him 00:06:12.439 --> 00:06:12.470 align:start position:0% Someone asked him 00:06:12.470 --> 00:06:13.551 align:start position:0% Someone asked him to <00:06:12.645>write <00:06:12.820>a <00:06:12.995>beautiful 00:06:13.551 --> 00:06:13.582 align:start position:0% to write a beautiful 00:06:13.582 --> 00:06:14.169 align:start position:0% to write a beautiful symphony <00:06:14.107>and 00:06:14.169 --> 00:06:14.200 align:start position:0% symphony and 00:06:14.200 --> 00:06:15.312 align:start position:0% symphony and synchronize <00:06:14.550>it <00:06:14.900>with <00:06:15.250>a 00:06:15.312 --> 00:06:15.343 align:start position:0% synchronize it with a 00:06:15.343 --> 00:06:16.678 align:start position:0% synchronize it with a video. <00:06:15.644>After <00:06:15.945>4 <00:06:16.246>hours, 00:06:16.678 --> 00:06:16.714 align:start position:0% video. After 4 hours, 00:06:16.714 --> 00:06:17.717 align:start position:0% video. After 4 hours, they <00:06:17.036>received <00:06:17.358>a <00:06:17.680>9- 00:06:17.717 --> 00:06:18.506 align:start position:0% they received a 9- minute, <00:06:18.470>38- 00:06:18.506 --> 00:06:19.581 align:start position:0% minute, 38- second <00:06:19.007>piece <00:06:19.508>in 00:06:19.581 --> 00:06:19.616 align:start position:0% second piece in 00:06:19.616 --> 00:06:20.978 align:start position:0% second piece in four <00:06:20.022>parts <00:06:20.428>with <00:06:20.834>a 00:06:20.978 --> 00:06:21.014 align:start position:0% four parts with a 00:06:21.014 --> 00:06:22.411 align:start position:0% four parts with a visualization <00:06:21.497>of <00:06:21.980>each 00:06:22.411 --> 00:06:22.447 align:start position:0% visualization of each 00:06:22.447 --> 00:06:23.905 align:start position:0% visualization of each note. <00:06:22.887>People <00:06:23.327>say <00:06:23.767>that 00:06:23.905 --> 00:06:23.939 align:start position:0% note. People say that 00:06:23.939 --> 00:06:25.061 align:start position:0% note. People say that this <00:06:24.054>is <00:06:24.169>no <00:06:24.284>longer <00:06:24.399>just <00:06:24.514>ordinary 00:06:25.061 --> 00:06:25.095 align:start position:0% this is no longer just ordinary 00:06:25.095 --> 00:06:25.979 align:start position:0% this is no longer just ordinary calculations, <00:06:25.344>this <00:06:25.593>is <00:06:25.842>a 00:06:25.979 --> 00:06:26.013 align:start position:0% calculations, this is a 00:06:26.013 --> 00:06:26.931 align:start position:0% calculations, this is a true <00:06:26.336>expression <00:06:26.659>of 00:06:26.931 --> 00:06:26.965 align:start position:0% true expression of 00:06:26.965 --> 00:06:27.833 align:start position:0% true expression of creativity. <00:06:27.767>And 00:06:27.833 --> 00:06:27.866 align:start position:0% creativity. And 00:06:27.866 --> 00:06:29.087 align:start position:0% creativity. And creating <00:06:28.559>symphonies 00:06:29.087 --> 00:06:29.120 align:start position:0% creating symphonies 00:06:29.120 --> 00:06:29.945 align:start position:0% creating symphonies from <00:06:29.219>abstract 00:06:29.945 --> 00:06:29.978 align:start position:0% from abstract 00:06:29.978 --> 00:06:31.265 align:start position:0% from abstract vector <00:06:30.605>mathematics 00:06:31.265 --> 00:06:32.553 align:start position:0% vector mathematics is, <00:06:31.683>yes, <00:06:32.101>something <00:06:32.519>significant. 00:06:32.553 --> 00:06:32.692 align:start position:0% is, yes, something significant. 00:06:32.692 --> 00:06:34.122 align:start position:0% is, yes, something significant. So <00:06:32.857>how <00:06:33.022>does <00:06:33.187>Anthropic <00:06:33.352>manage <00:06:33.517>to 00:06:34.122 --> 00:06:34.159 align:start position:0% So how does Anthropic manage to 00:06:34.159 --> 00:06:35.552 align:start position:0% So how does Anthropic manage to move <00:06:34.836>so <00:06:35.513>fast? 00:06:35.552 --> 00:06:35.652 align:start position:0% move so fast? 00:06:35.652 --> 00:06:36.502 align:start position:0% move so fast? The <00:06:35.951>answer <00:06:36.250>may 00:06:36.502 --> 00:06:36.534 align:start position:0% The answer may 00:06:36.534 --> 00:06:37.731 align:start position:0% The answer may lie <00:06:36.723>in <00:06:36.912>a <00:06:37.101>recent 00:06:37.731 --> 00:06:37.763 align:start position:0% lie in a recent 00:06:37.763 --> 00:06:38.488 align:start position:0% lie in a recent post <00:06:38.093>on <00:06:38.423>the 00:06:38.488 --> 00:06:38.519 align:start position:0% post on the 00:06:38.519 --> 00:06:39.559 align:start position:0% post on the engineering <00:06:39.244>blog 00:06:39.559 --> 00:06:39.590 align:start position:0% engineering blog 00:06:39.590 --> 00:06:40.820 align:start position:0% engineering blog Anthropic. <00:06:39.844>If <00:06:40.098>you <00:06:40.352>read 00:06:40.820 --> 00:06:40.859 align:start position:0% Anthropic. If you read 00:06:40.859 --> 00:06:42.371 align:start position:0% Anthropic. If you read carefully, <00:06:41.266>there <00:06:41.673>is 00:06:42.371 --> 00:06:42.410 align:start position:0% carefully, there is 00:06:42.410 --> 00:06:43.689 align:start position:0% carefully, there is a <00:06:42.650>lot <00:06:42.890>of <00:06:43.130>information <00:06:43.370>there, <00:06:43.610>and 00:06:43.689 --> 00:06:43.728 align:start position:0% a lot of information there, and 00:06:43.728 --> 00:06:44.465 align:start position:0% a lot of information there, and many 00:06:44.465 --> 00:06:44.503 align:start position:0% many 00:06:44.503 --> 00:06:46.016 align:start position:0% many have <00:06:44.748>become <00:06:44.993>convinced <00:06:45.238>that <00:06:45.483>Fable <00:06:45.728>5.5 <00:06:45.973>was 00:06:46.016 --> 00:06:46.054 align:start position:0% have become convinced that Fable 5.5 was 00:06:46.054 --> 00:06:47.024 align:start position:0% have become convinced that Fable 5.5 was most <00:06:46.985>likely 00:06:47.024 --> 00:06:47.063 align:start position:0% most likely 00:06:47.063 --> 00:06:48.668 align:start position:0% most likely written <00:06:47.340>by <00:06:47.617>the <00:06:47.894>AI ​​<00:06:48.171>itself. <00:06:48.448>We're 00:06:48.668 --> 00:06:48.704 align:start position:0% written by the AI ​​itself. We're 00:06:48.704 --> 00:06:50.108 align:start position:0% written by the AI ​​itself. We're still <00:06:48.947>reading <00:06:49.190>between <00:06:49.433>the <00:06:49.676>lines 00:06:50.108 --> 00:06:51.511 align:start position:0% still reading between the lines , <00:06:50.527>but <00:06:50.946>honestly, <00:06:51.365>that 00:06:51.511 --> 00:06:51.547 align:start position:0% , but honestly, that 00:06:51.547 --> 00:06:53.663 align:start position:0% , but honestly, that explains <00:06:52.055>the <00:06:52.563>pace. <00:06:53.071>And <00:06:53.579>it’s 00:06:53.663 --> 00:06:53.706 align:start position:0% explains the pace. And it’s 00:06:53.706 --> 00:06:55.188 align:start position:0% explains the pace. And it’s this <00:06:55.018>idea—AI 00:06:55.188 --> 00:06:55.230 align:start position:0% this idea—AI 00:06:55.230 --> 00:06:56.839 align:start position:0% this idea—AI building <00:06:55.653>AI—that’s <00:06:56.076>fueling 00:06:56.839 --> 00:06:56.882 align:start position:0% building AI—that’s fueling 00:06:56.882 --> 00:06:58.533 align:start position:0% building AI—that’s fueling the <00:06:57.252>next <00:06:57.622>rumor <00:06:57.992>about <00:06:58.362>Opus 00:06:58.533 --> 00:06:58.576 align:start position:0% the next rumor about Opus 00:06:58.576 --> 00:07:00.551 align:start position:0% the next rumor about Opus 6. <00:06:59.955>Julian 00:07:00.551 --> 00:07:00.589 align:start position:0% 6. Julian 00:07:00.589 --> 00:07:01.669 align:start position:0% 6. Julian Horsey’s <00:07:00.793>article <00:07:00.997>on <00:07:01.201>Geeky <00:07:01.405>Gadgets 00:07:01.669 --> 00:07:01.706 align:start position:0% Horsey’s article on Geeky Gadgets 00:07:01.706 --> 00:07:02.823 align:start position:0% Horsey’s article on Geeky Gadgets says <00:07:02.097>that <00:07:02.488>Anthropic 00:07:02.823 --> 00:07:02.860 align:start position:0% says that Anthropic 00:07:02.860 --> 00:07:03.940 align:start position:0% says that Anthropic is <00:07:03.046>preparing <00:07:03.232>Claude <00:07:03.418>Opus <00:07:03.604>6 <00:07:03.790>for 00:07:03.940 --> 00:07:03.978 align:start position:0% is preparing Claude Opus 6 for 00:07:03.978 --> 00:07:05.281 align:start position:0% is preparing Claude Opus 6 for launch <00:07:04.536>next 00:07:05.281 --> 00:07:05.318 align:start position:0% launch next 00:07:05.318 --> 00:07:06.473 align:start position:0% launch next week, <00:07:05.783>ostensibly <00:07:06.248>to 00:07:06.473 --> 00:07:06.510 align:start position:0% week, ostensibly to 00:07:06.510 --> 00:07:07.441 align:start position:0% week, ostensibly to compete <00:07:07.367>with 00:07:07.441 --> 00:07:07.478 align:start position:0% compete with 00:07:07.478 --> 00:07:08.707 align:start position:0% compete with OpenAI’s <00:07:07.788>project, <00:07:08.098>which <00:07:08.408>is 00:07:08.707 --> 00:07:08.745 align:start position:0% OpenAI’s project, which is 00:07:08.745 --> 00:07:09.639 align:start position:0% OpenAI’s project, which is called <00:07:09.601>Bell. 00:07:09.639 --> 00:07:09.739 align:start position:0% called Bell. 00:07:09.739 --> 00:07:11.023 align:start position:0% called Bell. The <00:07:10.343>argument <00:07:10.947>is 00:07:11.023 --> 00:07:11.061 align:start position:0% The argument is 00:07:11.061 --> 00:07:12.421 align:start position:0% The argument is that <00:07:11.476>Opus <00:07:11.891>5.5's <00:07:12.306>design 00:07:12.421 --> 00:07:12.459 align:start position:0% that Opus 5.5's design 00:07:12.459 --> 00:07:13.404 align:start position:0% that Opus 5.5's design and <00:07:12.648>3D 00:07:13.404 --> 00:07:13.441 align:start position:0% and 3D 00:07:13.441 --> 00:07:14.915 align:start position:0% and 3D modeling <00:07:14.310>suggest 00:07:14.915 --> 00:07:14.953 align:start position:0% modeling suggest 00:07:14.953 --> 00:07:15.973 align:start position:0% modeling suggest that <00:07:15.633>Anthropic 00:07:15.973 --> 00:07:16.011 align:start position:0% that Anthropic 00:07:16.011 --> 00:07:17.220 align:start position:0% that Anthropic is <00:07:16.177>stuck <00:07:16.343>in <00:07:16.509>a <00:07:16.675>cycle <00:07:16.841>of 00:07:17.220 --> 00:07:17.258 align:start position:0% is stuck in a cycle of 00:07:17.258 --> 00:07:18.165 align:start position:0% is stuck in a cycle of recursive 00:07:18.165 --> 00:07:18.202 align:start position:0% recursive 00:07:18.202 --> 00:07:19.525 align:start position:0% recursive self-improvement, 00:07:19.525 --> 00:07:19.563 align:start position:0% self-improvement, 00:07:19.563 --> 00:07:20.734 align:start position:0% self-improvement, so <00:07:19.780>Opus <00:07:19.997>6 <00:07:20.214>must <00:07:20.431>be 00:07:20.734 --> 00:07:20.772 align:start position:0% so Opus 6 must be 00:07:20.772 --> 00:07:22.298 align:start position:0% so Opus 6 must be close. <00:07:21.080>However, <00:07:21.388>there <00:07:21.696>is <00:07:22.004>one 00:07:22.298 --> 00:07:22.335 align:start position:0% close. However, there is one 00:07:22.335 --> 00:07:23.284 align:start position:0% close. However, there is one problem. <00:07:23.053>If 00:07:23.284 --> 00:07:23.313 align:start position:0% problem. If 00:07:23.313 --> 00:07:24.298 align:start position:0% problem. If you <00:07:23.515>trace <00:07:23.717>everything <00:07:23.919>back <00:07:24.121>to 00:07:24.298 --> 00:07:24.327 align:start position:0% you trace everything back to 00:07:24.327 --> 00:07:25.109 align:start position:0% you trace everything back to the <00:07:24.530>beginning, <00:07:24.733>the <00:07:24.936>whole 00:07:25.109 --> 00:07:25.138 align:start position:0% the beginning, the whole 00:07:25.138 --> 00:07:26.239 align:start position:0% the beginning, the whole story <00:07:25.447>comes <00:07:25.756>from <00:07:26.065>a 00:07:26.239 --> 00:07:26.268 align:start position:0% story comes from a 00:07:26.268 --> 00:07:27.224 align:start position:0% story comes from a single <00:07:26.659>article, <00:07:27.050>which 00:07:27.224 --> 00:07:27.253 align:start position:0% single article, which 00:07:27.253 --> 00:07:28.296 align:start position:0% single article, which itself <00:07:27.716>references <00:07:28.179>a 00:07:28.296 --> 00:07:28.325 align:start position:0% itself references a 00:07:28.325 --> 00:07:29.310 align:start position:0% itself references a source <00:07:28.962>called 00:07:29.310 --> 00:07:29.339 align:start position:0% source called 00:07:29.339 --> 00:07:30.754 align:start position:0% source called Universe <00:07:29.670>of <00:07:30.001>AI. <00:07:30.332>No 00:07:30.754 --> 00:07:30.789 align:start position:0% Universe of AI. No 00:07:30.789 --> 00:07:32.020 align:start position:0% Universe of AI. No model <00:07:31.739>card, 00:07:32.020 --> 00:07:32.056 align:start position:0% model card, 00:07:32.056 --> 00:07:33.006 align:start position:0% model card, release <00:07:32.513>date, <00:07:32.970>tests, 00:07:33.006 --> 00:07:33.041 align:start position:0% release date, tests, 00:07:33.041 --> 00:07:34.413 align:start position:0% release date, tests, number <00:07:33.375>of <00:07:33.709>parameters 00:07:34.413 --> 00:07:35.750 align:start position:0% number of parameters , <00:07:35.064>context <00:07:35.715>window, 00:07:35.750 --> 00:07:35.785 align:start position:0% , context window, 00:07:35.785 --> 00:07:37.087 align:start position:0% , context window, price, <00:07:35.975>and <00:07:36.165>not <00:07:36.355>a <00:07:36.545>single <00:07:36.735>word 00:07:37.087 --> 00:07:37.122 align:start position:0% price, and not a single word 00:07:37.122 --> 00:07:38.108 align:start position:0% price, and not a single word from <00:07:37.597>Anthropic <00:07:38.072>itself. 00:07:38.108 --> 00:07:39.052 align:start position:0% from Anthropic itself. 00:07:39.052 --> 00:07:40.133 align:start position:0% from Anthropic itself. During <00:07:39.350>the <00:07:39.648>day, <00:07:39.946>it 00:07:40.133 --> 00:07:40.164 align:start position:0% During the day, it 00:07:40.164 --> 00:07:41.277 align:start position:0% During the day, it still <00:07:40.658>spread <00:07:41.152>across 00:07:41.277 --> 00:07:41.308 align:start position:0% still spread across 00:07:41.308 --> 00:07:42.142 align:start position:0% still spread across aggregators <00:07:42.019>and 00:07:42.142 --> 00:07:42.173 align:start position:0% aggregators and 00:07:42.173 --> 00:07:43.348 align:start position:0% aggregators and social <00:07:43.317>feeds. 00:07:43.348 --> 00:07:43.556 align:start position:0% social feeds. 00:07:43.556 --> 00:07:44.631 align:start position:0% social feeds. Compare <00:07:43.888>this <00:07:44.220>to <00:07:44.552>the 00:07:44.631 --> 00:07:44.669 align:start position:0% Compare this to the 00:07:44.669 --> 00:07:45.629 align:start position:0% Compare this to the pre- 00:07:45.629 --> 00:07:45.668 align:start position:0% pre- 00:07:45.668 --> 00:07:47.165 align:start position:0% pre- release <00:07:45.850>rumors <00:07:46.032>of <00:07:46.214>the <00:07:46.396>Opus <00:07:46.578>5.5 <00:07:46.760>and <00:07:46.942>Haiku <00:07:47.124>5.5, 00:07:47.165 --> 00:07:47.204 align:start position:0% release rumors of the Opus 5.5 and Haiku 5.5, 00:07:47.204 --> 00:07:48.164 align:start position:0% release rumors of the Opus 5.5 and Haiku 5.5, which <00:07:47.338>at <00:07:47.472>least 00:07:48.164 --> 00:07:48.202 align:start position:0% which at least 00:07:48.202 --> 00:07:49.085 align:start position:0% which at least indicated 00:07:49.085 --> 00:07:49.124 align:start position:0% indicated 00:07:49.124 --> 00:07:50.545 align:start position:0% indicated internal <00:07:49.796>signals, <00:07:50.468>and 00:07:50.545 --> 00:07:50.583 align:start position:0% internal signals, and 00:07:50.583 --> 00:07:51.543 align:start position:0% internal signals, and both <00:07:51.082>models 00:07:51.543 --> 00:07:51.581 align:start position:0% both models 00:07:51.581 --> 00:07:52.926 align:start position:0% both models appeared <00:07:52.465>within 00:07:52.926 --> 00:07:52.964 align:start position:0% appeared within 00:07:52.964 --> 00:07:54.193 align:start position:0% appeared within days <00:07:53.040>of <00:07:53.116>each <00:07:53.192>other; <00:07:53.268>there <00:07:53.344>is <00:07:53.420>nothing <00:07:53.496>like <00:07:53.572>this <00:07:53.648>for <00:07:53.724>Opus <00:07:53.800>6 <00:07:53.876>yet 00:07:54.193 --> 00:07:55.114 align:start position:0% days of each other; there is nothing like this for Opus 6 yet 00:07:55.114 --> 00:07:55.153 align:start position:0% 00:07:55.153 --> 00:07:56.709 align:start position:0% . <00:07:56.199>Is 00:07:56.709 --> 00:07:56.752 align:start position:0% . Is 00:07:56.752 --> 00:07:58.409 align:start position:0% . Is Anthropic <00:07:56.932>working <00:07:57.112>on <00:07:57.292>a <00:07:57.472>successor 00:07:58.409 --> 00:07:59.592 align:start position:0% Anthropic working on a successor ? <00:07:58.688>Of <00:07:58.967>course, <00:07:59.246>every 00:07:59.592 --> 00:07:59.626 align:start position:0% ? Of course, every 00:07:59.626 --> 00:08:00.830 align:start position:0% ? Of course, every laboratory <00:08:00.021>is <00:08:00.416>always 00:08:00.830 --> 00:08:00.864 align:start position:0% laboratory is always 00:08:00.864 --> 00:08:02.206 align:start position:0% laboratory is always working, <00:08:01.500>but <00:08:02.136>“maybe” 00:08:02.206 --> 00:08:02.240 align:start position:0% working, but “maybe” 00:08:02.240 --> 00:08:03.547 align:start position:0% working, but “maybe” and <00:08:02.859>“next <00:08:03.478>week” 00:08:03.547 --> 00:08:05.109 align:start position:0% and “next week” are <00:08:03.845>two <00:08:04.143>different <00:08:04.441>things. <00:08:04.739>There <00:08:05.037>is 00:08:05.109 --> 00:08:05.143 align:start position:0% are two different things. There is 00:08:05.143 --> 00:08:06.010 align:start position:0% are two different things. There is also <00:08:05.334>a <00:08:05.525>strong 00:08:06.010 --> 00:08:06.044 align:start position:0% also a strong 00:08:06.044 --> 00:08:06.980 align:start position:0% also a strong argument <00:08:06.634>against 00:08:06.980 --> 00:08:07.015 align:start position:0% argument against 00:08:07.015 --> 00:08:08.313 align:start position:0% argument against rushing. <00:08:08.036>Anthropic's 00:08:08.313 --> 00:08:08.344 align:start position:0% rushing. Anthropic's 00:08:08.344 --> 00:08:09.146 align:start position:0% rushing. Anthropic's IPO <00:08:08.590>filing <00:08:08.836>contains <00:08:09.082>80 00:08:09.146 --> 00:08:09.177 align:start position:0% IPO filing contains 80 00:08:09.177 --> 00:08:10.257 align:start position:0% IPO filing contains 80 pages <00:08:09.701>of 00:08:10.257 --> 00:08:10.288 align:start position:0% pages of 00:08:10.288 --> 00:08:10.997 align:start position:0% pages of AI <00:08:10.503>risk <00:08:10.718>disclosures, <00:08:10.933>and 00:08:10.997 --> 00:08:11.028 align:start position:0% AI risk disclosures, and 00:08:11.028 --> 00:08:11.707 align:start position:0% AI risk disclosures, and the <00:08:11.583>company 00:08:11.707 --> 00:08:11.738 align:start position:0% the company 00:08:11.738 --> 00:08:12.848 align:start position:0% the company preparing <00:08:12.108>to <00:08:12.478>go 00:08:12.848 --> 00:08:12.879 align:start position:0% preparing to go 00:08:12.879 --> 00:08:14.051 align:start position:0% preparing to go public <00:08:13.403>wants <00:08:13.927>its 00:08:14.051 --> 00:08:14.082 align:start position:0% public wants its 00:08:14.082 --> 00:08:14.915 align:start position:0% public wants its release <00:08:14.282>schedule <00:08:14.482>to 00:08:14.915 --> 00:08:14.946 align:start position:0% release schedule to 00:08:14.946 --> 00:08:15.440 align:start position:0% release schedule to appear 00:08:15.440 --> 00:08:15.470 align:start position:0% appear 00:08:15.470 --> 00:08:16.674 align:start position:0% appear disciplined, <00:08:16.010>not <00:08:16.550>a 00:08:16.674 --> 00:08:16.704 align:start position:0% disciplined, not a 00:08:16.704 --> 00:08:17.784 align:start position:0% disciplined, not a panicked <00:08:17.753>rush 00:08:17.784 --> 00:08:17.815 align:start position:0% panicked rush 00:08:17.815 --> 00:08:18.710 align:start position:0% panicked rush to <00:08:18.031>outdo 00:08:18.710 --> 00:08:18.741 align:start position:0% to outdo 00:08:18.741 --> 00:08:19.790 align:start position:0% to outdo competitors' <00:08:19.759>rumors. 00:08:19.790 --> 00:08:20.007 align:start position:0% competitors' rumors. 00:08:20.007 --> 00:08:21.297 align:start position:0% competitors' rumors. Observers <00:08:20.834>are 00:08:21.297 --> 00:08:21.330 align:start position:0% Observers are 00:08:21.330 --> 00:08:22.587 align:start position:0% Observers are betting <00:08:21.649>that <00:08:21.968>Anthropic <00:08:22.287>will 00:08:22.587 --> 00:08:22.620 align:start position:0% betting that Anthropic will 00:08:22.620 --> 00:08:23.679 align:start position:0% betting that Anthropic will simply <00:08:23.132>keep <00:08:23.644>quiet 00:08:23.679 --> 00:08:23.712 align:start position:0% simply keep quiet 00:08:23.712 --> 00:08:24.704 align:start position:0% simply keep quiet rather <00:08:24.506>than 00:08:24.704 --> 00:08:24.737 align:start position:0% rather than 00:08:24.737 --> 00:08:25.829 align:start position:0% rather than deny <00:08:25.079>it, <00:08:25.421>and <00:08:25.763>in 00:08:25.829 --> 00:08:25.862 align:start position:0% deny it, and in 00:08:25.862 --> 00:08:26.788 align:start position:0% deny it, and in the <00:08:26.159>meantime, <00:08:26.456>some 00:08:26.788 --> 00:08:26.821 align:start position:0% the meantime, some 00:08:26.821 --> 00:08:28.111 align:start position:0% the meantime, some corporate <00:08:27.648>buyers 00:08:28.111 --> 00:08:28.144 align:start position:0% corporate buyers 00:08:28.144 --> 00:08:29.170 align:start position:0% corporate buyers may <00:08:28.287>hold <00:08:28.430>off <00:08:28.573>on 00:08:29.170 --> 00:08:29.203 align:start position:0% may hold off on 00:08:29.203 --> 00:08:29.865 align:start position:0% may hold off on signing 00:08:29.865 --> 00:08:29.898 align:start position:0% signing 00:08:29.898 --> 00:08:30.923 align:start position:0% signing contracts <00:08:30.394>with <00:08:30.890>Claude 00:08:30.923 --> 00:08:30.956 align:start position:0% contracts with Claude 00:08:30.956 --> 00:08:32.213 align:start position:0% contracts with Claude until <00:08:31.237>the <00:08:31.518>situation <00:08:31.799>with <00:08:32.080>Bell 00:08:32.213 --> 00:08:32.247 align:start position:0% until the situation with Bell 00:08:32.247 --> 00:08:33.471 align:start position:0% until the situation with Bell becomes <00:08:32.842>less <00:08:33.437>murky. 00:08:33.471 --> 00:08:33.845 align:start position:0% becomes less murky. 00:08:33.845 --> 00:08:34.972 align:start position:0% becomes less murky. One <00:08:34.392>site, <00:08:34.939>key.ai, 00:08:34.972 --> 00:08:35.004 align:start position:0% One site, key.ai, 00:08:35.004 --> 00:08:35.939 align:start position:0% One site, key.ai, describes <00:08:35.205>OpenAI <00:08:35.406>Bell <00:08:35.607>as <00:08:35.808>an 00:08:35.939 --> 00:08:35.971 align:start position:0% describes OpenAI Bell as an 00:08:35.971 --> 00:08:36.873 align:start position:0% describes OpenAI Bell as an unconfirmed 00:08:36.873 --> 00:08:36.905 align:start position:0% unconfirmed 00:08:36.905 --> 00:08:37.872 align:start position:0% unconfirmed codename <00:08:37.292>for <00:08:37.679>the 00:08:37.872 --> 00:08:37.904 align:start position:0% codename for the 00:08:37.904 --> 00:08:39.032 align:start position:0% codename for the pre- 00:08:39.032 --> 00:08:39.064 align:start position:0% pre- 00:08:39.064 --> 00:08:40.288 align:start position:0% pre- training <00:08:39.321>stage <00:08:39.578>or <00:08:39.835>base 00:08:40.288 --> 00:08:40.321 align:start position:0% training stage or base 00:08:40.321 --> 00:08:41.223 align:start position:0% training stage or base model, <00:08:40.772>with 00:08:41.223 --> 00:08:41.255 align:start position:0% model, with 00:08:41.255 --> 00:08:42.479 align:start position:0% model, with the <00:08:41.851>community <00:08:42.447>claiming 00:08:42.479 --> 00:08:42.512 align:start position:0% the community claiming 00:08:42.512 --> 00:08:43.607 align:start position:0% the community claiming it <00:08:42.855>has <00:08:43.198>over <00:08:43.541>10 00:08:43.607 --> 00:08:43.639 align:start position:0% it has over 10 00:08:43.639 --> 00:08:44.284 align:start position:0% it has over 10 trillion 00:08:44.284 --> 00:08:44.316 align:start position:0% trillion 00:08:44.316 --> 00:08:45.921 align:start position:0% trillion parameters. <00:08:45.636>Another 00:08:45.921 --> 00:08:45.957 align:start position:0% parameters. Another 00:08:45.957 --> 00:08:47.098 align:start position:0% parameters. Another version <00:08:46.474>says <00:08:46.991>100 00:08:47.098 --> 00:08:47.134 align:start position:0% version says 100 00:08:47.134 --> 00:08:48.845 align:start position:0% version says 100 trillion. <00:08:47.655>This <00:08:48.176>is <00:08:48.697>a 00:08:48.845 --> 00:08:48.882 align:start position:0% trillion. This is a 00:08:48.882 --> 00:08:50.324 align:start position:0% trillion. This is a hundredfold <00:08:49.529>difference, <00:08:50.176>which 00:08:50.324 --> 00:08:51.765 align:start position:0% hundredfold difference, which actually <00:08:50.933>says <00:08:51.542>it 00:08:51.765 --> 00:08:51.802 align:start position:0% actually says it 00:08:51.802 --> 00:08:53.611 align:start position:0% actually says it all. <00:08:52.038>AX <00:08:52.274>Brief <00:08:52.510>lumped <00:08:52.746>this <00:08:52.982>in <00:08:53.218>with <00:08:53.454>a 00:08:53.611 --> 00:08:53.648 align:start position:0% all. AX Brief lumped this in with a 00:08:53.648 --> 00:08:54.365 align:start position:0% all. AX Brief lumped this in with a bunch <00:08:53.818>of <00:08:53.988>other 00:08:54.365 --> 00:08:54.402 align:start position:0% bunch of other 00:08:54.402 --> 00:08:55.836 align:start position:0% bunch of other unverified <00:08:55.798>rumors. 00:08:55.836 --> 00:08:56.274 align:start position:0% unverified rumors. 00:08:56.274 --> 00:08:57.442 align:start position:0% unverified rumors. OpenAI <00:08:56.478>has <00:08:56.682>not <00:08:56.886>mentioned <00:08:57.090>this <00:08:57.294>anywhere 00:08:57.442 --> 00:08:57.479 align:start position:0% OpenAI has not mentioned this anywhere 00:08:57.479 --> 00:08:58.903 align:start position:0% OpenAI has not mentioned this anywhere , <00:08:57.880>and <00:08:58.281>it <00:08:58.682>has 00:08:58.903 --> 00:08:58.939 align:start position:0% , and it has 00:08:58.939 --> 00:09:00.254 align:start position:0% , and it has nothing <00:08:59.328>to <00:08:59.717>do <00:09:00.106>with 00:09:00.254 --> 00:09:00.291 align:start position:0% nothing to do with 00:09:00.291 --> 00:09:01.459 align:start position:0% nothing to do with OpenAI's <00:09:00.820>Operation <00:09:01.349>Nobel— 00:09:01.459 --> 00:09:02.080 align:start position:0% OpenAI's Operation Nobel— a <00:09:01.660>report <00:09:01.861>on 00:09:02.080 --> 00:09:02.117 align:start position:0% a report on 00:09:02.117 --> 00:09:03.212 align:start position:0% a report on disinformation 00:09:03.212 --> 00:09:03.249 align:start position:0% disinformation 00:09:03.249 --> 00:09:04.198 align:start position:0% disinformation campaigns <00:09:03.906>that 00:09:04.198 --> 00:09:04.235 align:start position:0% campaigns that 00:09:04.235 --> 00:09:05.477 align:start position:0% campaigns that coincidentally <00:09:04.709>has <00:09:05.183>the 00:09:05.477 --> 00:09:05.513 align:start position:0% coincidentally has the 00:09:05.513 --> 00:09:06.970 align:start position:0% coincidentally has the same <00:09:05.955>name. <00:09:06.397>So, <00:09:06.839>neither 00:09:06.970 --> 00:09:07.002 align:start position:0% same name. So, neither 00:09:07.002 --> 00:09:07.755 align:start position:0% same name. So, neither Fable <00:09:07.157>nor <00:09:07.312>Opus <00:09:07.467>6 <00:09:07.622>are 00:09:07.755 --> 00:09:07.788 align:start position:0% Fable nor Opus 6 are 00:09:07.788 --> 00:09:08.933 align:start position:0% Fable nor Opus 6 are officially <00:09:08.344>out <00:09:08.900>yet, 00:09:08.933 --> 00:09:08.965 align:start position:0% officially out yet, 00:09:08.965 --> 00:09:09.849 align:start position:0% officially out yet, but <00:09:09.243>Anthropic <00:09:09.521>just 00:09:09.849 --> 00:09:09.882 align:start position:0% but Anthropic just 00:09:09.882 --> 00:09:11.027 align:start position:0% but Anthropic just released <00:09:10.830>something 00:09:11.027 --> 00:09:11.059 align:start position:0% released something 00:09:11.059 --> 00:09:11.616 align:start position:0% released something you <00:09:11.223>can 00:09:11.616 --> 00:09:11.648 align:start position:0% you can 00:09:11.648 --> 00:09:12.663 align:start position:0% you can take <00:09:12.057>advantage <00:09:12.466>of 00:09:12.663 --> 00:09:12.695 align:start position:0% take advantage of 00:09:12.695 --> 00:09:13.771 align:start position:0% take advantage of today. 00:09:13.771 --> 00:09:13.801 align:start position:0% today. 00:09:13.801 --> 00:09:14.881 align:start position:0% today. Fast <00:09:14.101>mode <00:09:14.401>for <00:09:14.701>Claude 00:09:14.881 --> 00:09:14.911 align:start position:0% Fast mode for Claude 00:09:14.911 --> 00:09:15.691 align:start position:0% Fast mode for Claude Opus <00:09:15.043>5.5 <00:09:15.175>has <00:09:15.307>moved <00:09:15.439>out <00:09:15.571>of 00:09:15.691 --> 00:09:15.721 align:start position:0% Opus 5.5 has moved out of 00:09:15.721 --> 00:09:16.560 align:start position:0% Opus 5.5 has moved out of research 00:09:16.560 --> 00:09:16.590 align:start position:0% research 00:09:16.590 --> 00:09:17.340 align:start position:0% research preview, 00:09:17.340 --> 00:09:17.370 align:start position:0% preview, 00:09:17.370 --> 00:09:18.539 align:start position:0% preview, waitlists <00:09:17.580>have <00:09:17.790>been <00:09:18.000>removed 00:09:18.539 --> 00:09:19.649 align:start position:0% waitlists have been removed , <00:09:18.725>and <00:09:18.911>it <00:09:19.097>is <00:09:19.283>open <00:09:19.469>to 00:09:19.649 --> 00:09:19.679 align:start position:0% , and it is open to 00:09:19.679 --> 00:09:20.759 align:start position:0% , and it is open to every <00:09:20.729>developer. 00:09:20.759 --> 00:09:20.859 align:start position:0% every developer. 00:09:20.859 --> 00:09:21.977 align:start position:0% every developer. No 00:09:21.977 --> 00:09:22.013 align:start position:0% No 00:09:22.013 --> 00:09:23.059 align:start position:0% No performance <00:09:23.023>degradation, 00:09:23.059 --> 00:09:23.095 align:start position:0% performance degradation, 00:09:23.095 --> 00:09:24.141 align:start position:0% performance degradation, no 00:09:24.141 --> 00:09:24.177 align:start position:0% no 00:09:24.177 --> 00:09:25.440 align:start position:0% no feature <00:09:24.402>limitations, <00:09:24.627>one <00:09:24.852>line <00:09:25.077>of 00:09:25.440 --> 00:09:25.476 align:start position:0% feature limitations, one line of 00:09:25.476 --> 00:09:26.558 align:start position:0% feature limitations, one line of code—up <00:09:25.837>to <00:09:26.198>2.5x 00:09:26.558 --> 00:09:26.595 align:start position:0% code—up to 2.5x 00:09:26.595 --> 00:09:27.462 align:start position:0% code—up to 2.5x faster. <00:09:27.164>One 00:09:27.462 --> 00:09:27.499 align:start position:0% faster. One 00:09:27.499 --> 00:09:28.840 align:start position:0% faster. One developer <00:09:28.095>called <00:09:28.691>Opus 00:09:28.840 --> 00:09:28.877 align:start position:0% developer called Opus 00:09:28.877 --> 00:09:30.107 align:start position:0% developer called Opus 5.5 <00:09:30.070>“divine,” 00:09:30.107 --> 00:09:30.144 align:start position:0% 5.5 “divine,” 00:09:30.144 --> 00:09:31.299 align:start position:0% 5.5 “divine,” but <00:09:30.231>the <00:09:30.318>most <00:09:30.405>telling 00:09:31.299 --> 00:09:31.336 align:start position:0% but the most telling 00:09:31.336 --> 00:09:32.678 align:start position:0% but the most telling comment <00:09:31.988>was <00:09:32.640>that 00:09:32.678 --> 00:09:32.715 align:start position:0% comment was that 00:09:32.715 --> 00:09:33.758 align:start position:0% comment was that for <00:09:33.162>agent 00:09:33.758 --> 00:09:33.795 align:start position:0% for agent 00:09:33.795 --> 00:09:35.025 align:start position:0% for agent cycles, <00:09:34.167>P95 <00:09:34.539>latency <00:09:34.911>is 00:09:35.025 --> 00:09:35.062 align:start position:0% cycles, P95 latency is 00:09:35.062 --> 00:09:36.515 align:start position:0% cycles, P95 latency is more <00:09:35.350>important <00:09:35.638>than <00:09:35.926>anything <00:09:36.214>else 00:09:36.515 --> 00:09:37.938 align:start position:0% more important than anything else . <00:09:36.943>Anthropic <00:09:37.371>claims <00:09:37.799>that 00:09:37.938 --> 00:09:37.973 align:start position:0% . Anthropic claims that 00:09:37.973 --> 00:09:39.273 align:start position:0% . Anthropic claims that this <00:09:38.178>is <00:09:38.383>the <00:09:38.588>exact <00:09:38.793>same <00:09:38.998>weight 00:09:39.273 --> 00:09:39.307 align:start position:0% this is the exact same weight 00:09:39.307 --> 00:09:40.437 align:start position:0% this is the exact same weight and <00:09:39.661>intelligence <00:09:40.015>as <00:09:40.369>the 00:09:40.437 --> 00:09:40.471 align:start position:0% and intelligence as the 00:09:40.471 --> 00:09:41.704 align:start position:0% and intelligence as the standard <00:09:41.669>version. 00:09:41.704 --> 00:09:41.869 align:start position:0% standard version. 00:09:41.869 --> 00:09:42.697 align:start position:0% standard version. The acceleration 00:09:42.697 --> 00:09:42.735 align:start position:0% The acceleration 00:09:42.735 --> 00:09:43.902 align:start position:0% The acceleration is <00:09:43.168>achieved <00:09:43.601>purely 00:09:43.902 --> 00:09:43.940 align:start position:0% is achieved purely 00:09:43.940 --> 00:09:45.333 align:start position:0% is achieved purely through <00:09:44.128>optimization <00:09:44.316>of <00:09:44.504>the 00:09:45.333 --> 00:09:45.371 align:start position:0% through optimization of the 00:09:45.371 --> 00:09:46.312 align:start position:0% through optimization of the underlying <00:09:45.766>API <00:09:46.161>and 00:09:46.312 --> 00:09:46.350 align:start position:0% underlying API and 00:09:46.350 --> 00:09:47.253 align:start position:0% underlying API and 00:09:47.253 --> 00:09:47.291 align:start position:0% 00:09:47.291 --> 00:09:48.760 align:start position:0% inference <00:09:47.579>configuration, <00:09:47.867>which <00:09:48.155>increases 00:09:48.760 --> 00:09:48.797 align:start position:0% inference configuration, which increases 00:09:48.797 --> 00:09:50.115 align:start position:0% inference configuration, which increases the <00:09:49.035>number <00:09:49.273>of <00:09:49.511>output 00:09:50.115 --> 00:09:50.153 align:start position:0% the number of output 00:09:50.153 --> 00:09:51.622 align:start position:0% the number of output tokens <00:09:50.605>per <00:09:51.057>second <00:09:51.509>(OTF 00:09:51.622 --> 00:09:51.659 align:start position:0% tokens per second (OTF 00:09:51.659 --> 00:09:52.625 align:start position:0% tokens per second (OTF PS) <00:09:51.829>by <00:09:51.999>2.5 <00:09:52.169>times. <00:09:52.339>This <00:09:52.509>is 00:09:52.625 --> 00:09:52.653 align:start position:0% PS) by 2.5 times. This is 00:09:52.653 --> 00:09:53.717 align:start position:0% PS) by 2.5 times. This is aimed <00:09:53.017>at <00:09:53.381>rapid 00:09:53.717 --> 00:09:53.745 align:start position:0% aimed at rapid 00:09:53.745 --> 00:09:54.809 align:start position:0% aimed at rapid iterative <00:09:54.781>development, 00:09:54.809 --> 00:09:55.593 align:start position:0% iterative development, 00:09:55.593 --> 00:09:55.621 align:start position:0% 00:09:55.621 --> 00:09:56.518 align:start position:0% real-time <00:09:56.013>debugging, <00:09:56.405>and 00:09:56.518 --> 00:09:56.546 align:start position:0% real-time debugging, and 00:09:56.546 --> 00:09:57.610 align:start position:0% real-time debugging, and urgent <00:09:57.050>tasks <00:09:57.554>with 00:09:57.610 --> 00:09:57.638 align:start position:0% urgent tasks with 00:09:57.638 --> 00:09:58.730 align:start position:0% urgent tasks with tight <00:09:58.170>deadlines 00:09:58.730 --> 00:09:59.809 align:start position:0% tight deadlines . <00:09:59.119>It <00:09:59.508>currently 00:09:59.809 --> 00:09:59.859 align:start position:0% . It currently 00:09:59.859 --> 00:10:01.710 align:start position:0% . It currently only <00:10:00.459>supports <00:10:01.059>Opus <00:10:01.659>5.5. 00:10:01.710 --> 00:10:01.810 align:start position:0% only supports Opus 5.5. 00:10:01.810 --> 00:10:05.475 align:start position:0% only supports Opus 5.5. Opus <00:10:02.197>5 <00:10:02.584>and <00:10:02.971>Opus <00:10:03.358>4.8. <00:10:03.745>Not <00:10:04.132>a <00:10:04.519>Sonnet, <00:10:04.906>not <00:10:05.293>a 00:10:05.475 --> 00:10:05.519 align:start position:0% Opus 5 and Opus 4.8. Not a Sonnet, not a 00:10:05.519 --> 00:10:07.156 align:start position:0% Opus 5 and Opus 4.8. Not a Sonnet, not a Haiku, <00:10:06.049>or <00:10:06.579>anything <00:10:07.109>else. 00:10:07.156 --> 00:10:07.626 align:start position:0% Haiku, or anything else. 00:10:07.626 --> 00:10:09.361 align:start position:0% Haiku, or anything else. And <00:10:07.966>the <00:10:08.306>older <00:10:08.646>fast 00:10:09.361 --> 00:10:09.412 align:start position:0% And the older fast 00:10:09.412 --> 00:10:10.740 align:start position:0% And the older fast mode <00:10:09.752>Opus <00:10:10.092>4.8 <00:10:10.432>was 00:10:10.740 --> 00:10:10.791 align:start position:0% mode Opus 4.8 was 00:10:10.791 --> 00:10:12.220 align:start position:0% mode Opus 4.8 was discontinued <00:10:11.454>back <00:10:12.117>in 00:10:12.220 --> 00:10:12.271 align:start position:0% discontinued back in 00:10:12.271 --> 00:10:13.650 align:start position:0% discontinued back in June <00:10:13.598>2026. 00:10:13.650 --> 00:10:13.750 align:start position:0% June 2026. 00:10:13.750 --> 00:10:14.716 align:start position:0% June 2026. Faster <00:10:14.183>usually 00:10:14.716 --> 00:10:14.750 align:start position:0% Faster usually 00:10:14.750 --> 00:10:15.983 align:start position:0% Faster usually means <00:10:15.094>more <00:10:15.438>expensive, <00:10:15.782>but 00:10:15.983 --> 00:10:16.017 align:start position:0% means more expensive, but 00:10:16.017 --> 00:10:17.305 align:start position:0% means more expensive, but here <00:10:16.267>it's <00:10:16.517>the <00:10:16.767>opposite. <00:10:17.017>Opus <00:10:17.267>5 00:10:17.305 --> 00:10:17.341 align:start position:0% here it's the opposite. Opus 5 00:10:17.341 --> 00:10:18.373 align:start position:0% here it's the opposite. Opus 5 and <00:10:17.488>4.8 <00:10:17.635>in <00:10:17.782>fast 00:10:18.373 --> 00:10:18.410 align:start position:0% and 4.8 in fast 00:10:18.410 --> 00:10:19.589 align:start position:0% and 4.8 in fast mode <00:10:19.515>cost 00:10:19.589 --> 00:10:19.626 align:start position:0% mode cost 00:10:19.626 --> 00:10:20.989 align:start position:0% mode cost $10 <00:10:20.031>for <00:10:20.436>inbound <00:10:20.841>and 00:10:20.989 --> 00:10:21.026 align:start position:0% $10 for inbound and 00:10:21.026 --> 00:10:22.390 align:start position:0% $10 for inbound and $50 <00:10:21.450>for <00:10:21.874>outbound 00:10:22.390 --> 00:10:22.427 align:start position:0% $50 for outbound 00:10:22.427 --> 00:10:23.903 align:start position:0% $50 for outbound tokens <00:10:22.865>per <00:10:23.303>million. <00:10:23.741>Opus 00:10:23.903 --> 00:10:23.943 align:start position:0% tokens per million. Opus 00:10:23.943 --> 00:10:25.388 align:start position:0% tokens per million. Opus 5.5 <00:10:24.264>in <00:10:24.585>fast <00:10:24.906>mode 00:10:25.388 --> 00:10:25.428 align:start position:0% 5.5 in fast mode 00:10:25.428 --> 00:10:26.913 align:start position:0% 5.5 in fast mode costs <00:10:25.735>$8 <00:10:26.042>and <00:10:26.349>$40 00:10:26.913 --> 00:10:26.953 align:start position:0% costs $8 and $40 00:10:26.953 --> 00:10:28.185 align:start position:0% costs $8 and $40 respectively. <00:10:27.553>So, <00:10:28.153>the 00:10:28.185 --> 00:10:28.217 align:start position:0% respectively. So, the 00:10:28.217 --> 00:10:29.407 align:start position:0% respectively. So, the most <00:10:28.377>powerful <00:10:28.537>model <00:10:28.697>is <00:10:28.857>the <00:10:29.017>cheapest 00:10:29.407 --> 00:10:30.351 align:start position:0% most powerful model is the cheapest . <00:10:29.843>There <00:10:30.279>are 00:10:30.351 --> 00:10:30.386 align:start position:0% . There are 00:10:30.386 --> 00:10:31.384 align:start position:0% . There are several <00:10:31.348>nuances. 00:10:31.384 --> 00:10:31.484 align:start position:0% several nuances. 00:10:31.484 --> 00:10:32.214 align:start position:0% several nuances. Fast <00:10:31.780>mode <00:10:32.076>has 00:10:32.214 --> 00:10:32.237 align:start position:0% Fast mode has 00:10:32.237 --> 00:10:32.807 align:start position:0% Fast mode has its <00:10:32.579>own 00:10:32.807 --> 00:10:32.830 align:start position:0% its own 00:10:32.830 --> 00:10:33.651 align:start position:0% its own request <00:10:33.012>limit, <00:10:33.194>so <00:10:33.376>it <00:10:33.558>doesn't 00:10:33.651 --> 00:10:33.674 align:start position:0% request limit, so it doesn't 00:10:33.674 --> 00:10:34.359 align:start position:0% request limit, so it doesn't compete <00:10:33.902>with <00:10:34.130>your 00:10:34.359 --> 00:10:34.382 align:start position:0% compete with your 00:10:34.382 --> 00:10:34.793 align:start position:0% compete with your normal 00:10:34.793 --> 00:10:34.815 align:start position:0% normal 00:10:34.815 --> 00:10:35.755 align:start position:0% normal usage. <00:10:35.204>And <00:10:35.593>if 00:10:35.755 --> 00:10:35.776 align:start position:0% usage. And if 00:10:35.776 --> 00:10:36.528 align:start position:0% usage. And if you <00:10:36.020>reach <00:10:36.264>the <00:10:36.508>limit, 00:10:36.528 --> 00:10:36.548 align:start position:0% you reach the limit, 00:10:36.548 --> 00:10:36.935 align:start position:0% you reach the limit, it <00:10:36.691>smoothly 00:10:36.935 --> 00:10:36.955 align:start position:0% it smoothly 00:10:36.955 --> 00:10:37.463 align:start position:0% it smoothly switches <00:10:37.382>to 00:10:37.463 --> 00:10:37.484 align:start position:0% switches to 00:10:37.484 --> 00:10:38.277 align:start position:0% switches to standard <00:10:37.911>speed 00:10:38.277 --> 00:10:38.297 align:start position:0% standard speed 00:10:38.297 --> 00:10:38.927 align:start position:0% standard speed instead <00:10:38.805>of 00:10:38.927 --> 00:10:38.948 align:start position:0% instead of 00:10:38.948 --> 00:10:39.599 align:start position:0% instead of giving <00:10:39.158>an <00:10:39.368>error. <00:10:39.578>The 00:10:39.599 --> 00:10:39.699 align:start position:0% giving an error. The 00:10:39.699 --> 00:10:40.537 align:start position:0% giving an error. The lightning <00:10:39.874>bolt <00:10:40.049>icon 00:10:40.537 --> 00:10:40.564 align:start position:0% lightning bolt icon 00:10:40.564 --> 00:10:41.457 align:start position:0% lightning bolt icon turns <00:10:40.929>gray <00:10:41.294>during 00:10:41.457 --> 00:10:41.484 align:start position:0% turns gray during 00:10:41.484 --> 00:10:42.611 align:start position:0% turns gray during cooling. <00:10:41.756>And <00:10:42.028>then <00:10:42.300>the 00:10:42.611 --> 00:10:42.642 align:start position:0% cooling. And then the 00:10:42.642 --> 00:10:43.605 align:start position:0% cooling. And then the 2.5x <00:10:42.875>acceleration <00:10:43.108>mode 00:10:43.605 --> 00:10:43.636 align:start position:0% 2.5x acceleration mode 00:10:43.636 --> 00:10:44.320 align:start position:0% 2.5x acceleration mode 00:10:44.320 --> 00:10:44.351 align:start position:0% 00:10:44.351 --> 00:10:45.097 align:start position:0% returns 00:10:45.097 --> 00:10:45.128 align:start position:0% returns 00:10:45.128 --> 00:10:46.287 align:start position:0% returns automatically. <00:10:45.506>The <00:10:45.884>same <00:10:46.262>applies 00:10:46.287 --> 00:10:46.312 align:start position:0% automatically. The same applies 00:10:46.312 --> 00:10:47.001 align:start position:0% automatically. The same applies if <00:10:46.459>your <00:10:46.606>usage <00:10:46.753>quota 00:10:47.001 --> 00:10:47.026 align:start position:0% if your usage quota 00:10:47.026 --> 00:10:47.617 align:start position:0% if your usage quota 00:10:47.617 --> 00:10:47.641 align:start position:0% 00:10:47.641 --> 00:10:48.602 align:start position:0% runs <00:10:47.764>out <00:10:47.887>in <00:10:48.010>the <00:10:48.133>middle <00:10:48.256>of 00:10:48.602 --> 00:10:48.626 align:start position:0% runs out in the middle of 00:10:48.626 --> 00:10:49.119 align:start position:0% runs out in the middle of a <00:10:48.774>session: <00:10:48.922>it 00:10:49.119 --> 00:10:49.143 align:start position:0% a session: it 00:10:49.143 --> 00:10:49.759 align:start position:0% a session: it is <00:10:49.315>reduced <00:10:49.487>to <00:10:49.659>the 00:10:49.759 --> 00:10:49.784 align:start position:0% is reduced to the 00:10:49.784 --> 00:10:50.325 align:start position:0% is reduced to the standard 00:10:50.325 --> 00:10:50.350 align:start position:0% standard 00:10:50.350 --> 00:10:51.113 align:start position:0% standard speed, <00:10:51.015>as 00:10:51.113 --> 00:10:51.138 align:start position:0% speed, as 00:10:51.138 --> 00:10:52.074 align:start position:0% speed, as reported <00:10:51.441>by <00:10:51.744>the <00:10:52.047>terminal. 00:10:52.074 --> 00:10:52.174 align:start position:0% reported by the terminal. 00:10:52.174 --> 00:10:53.280 align:start position:0% reported by the terminal. So <00:10:52.865>no 00:10:53.280 --> 00:10:53.315 align:start position:0% So no 00:10:53.315 --> 00:10:54.552 align:start position:0% So no job <00:10:53.541>loss. <00:10:53.767>A <00:10:53.993>bigger <00:10:54.219>pitfall 00:10:54.552 --> 00:10:54.580 align:start position:0% job loss. A bigger pitfall 00:10:54.580 --> 00:10:55.240 align:start position:0% job loss. A bigger pitfall is <00:10:54.827>that <00:10:55.074>the 00:10:55.240 --> 00:10:55.267 align:start position:0% is that the 00:10:55.267 --> 00:10:55.927 align:start position:0% is that the prompt <00:10:55.468>cache <00:10:55.669>is <00:10:55.870>not 00:10:55.927 --> 00:10:55.955 align:start position:0% prompt cache is not 00:10:55.955 --> 00:10:56.588 align:start position:0% prompt cache is not shared <00:10:56.423>between 00:10:56.588 --> 00:10:56.615 align:start position:0% shared between 00:10:56.615 --> 00:10:57.193 align:start position:0% shared between fast <00:10:57.083>and 00:10:57.193 --> 00:10:57.220 align:start position:0% fast and 00:10:57.220 --> 00:10:58.293 align:start position:0% fast and standard <00:10:57.908>modes 00:10:58.293 --> 00:10:58.907 align:start position:0% standard modes . <00:10:58.478>Therefore, <00:10:58.663>Anthropic 00:10:58.907 --> 00:10:58.934 align:start position:0% . Therefore, Anthropic 00:10:58.934 --> 00:10:59.880 align:start position:0% . Therefore, Anthropic recommends <00:10:59.501>enabling 00:10:59.880 --> 00:10:59.907 align:start position:0% recommends enabling 00:10:59.907 --> 00:11:00.772 align:start position:0% recommends enabling fast <00:11:00.258>mode <00:11:00.609>from 00:11:00.772 --> 00:11:00.799 align:start position:0% fast mode from 00:11:00.799 --> 00:11:01.555 align:start position:0% fast mode from the <00:11:01.528>start, 00:11:01.555 --> 00:11:01.582 align:start position:0% the start, 00:11:01.582 --> 00:11:02.636 align:start position:0% the start, as <00:11:02.042>switching 00:11:02.636 --> 00:11:02.663 align:start position:0% as switching 00:11:02.663 --> 00:11:03.447 align:start position:0% as switching in <00:11:02.744>the <00:11:02.825>middle <00:11:02.906>of <00:11:02.987>the <00:11:03.068>process 00:11:03.447 --> 00:11:03.474 align:start position:0% in the middle of the process 00:11:03.474 --> 00:11:04.204 align:start position:0% in the middle of the process means <00:11:03.609>paying <00:11:03.744>for <00:11:03.879>a 00:11:04.204 --> 00:11:04.231 align:start position:0% means paying for a 00:11:04.231 --> 00:11:04.609 align:start position:0% means paying for a fresh, 00:11:04.609 --> 00:11:04.636 align:start position:0% fresh, 00:11:04.636 --> 00:11:05.393 align:start position:0% fresh, uncached 00:11:05.393 --> 00:11:05.420 align:start position:0% uncached 00:11:05.420 --> 00:11:06.540 align:start position:0% uncached incoming <00:11:05.875>request. <00:11:06.330>However 00:11:06.540 --> 00:11:07.481 align:start position:0% incoming request. However , <00:11:06.723>you <00:11:06.906>are <00:11:07.089>only <00:11:07.272>charged 00:11:07.481 --> 00:11:07.507 align:start position:0% , you are only charged 00:11:07.507 --> 00:11:08.449 align:start position:0% , you are only charged once <00:11:07.965>per <00:11:08.423>call, 00:11:08.449 --> 00:11:08.475 align:start position:0% once per call, 00:11:08.475 --> 00:11:09.390 align:start position:0% once per call, no <00:11:09.364>matter 00:11:09.390 --> 00:11:09.416 align:start position:0% no matter 00:11:09.416 --> 00:11:10.201 align:start position:0% no matter how <00:11:09.642>many <00:11:09.868>times <00:11:10.094>you 00:11:10.201 --> 00:11:10.227 align:start position:0% how many times you 00:11:10.227 --> 00:11:11.195 align:start position:0% how many times you switch <00:11:10.933>after 00:11:11.195 --> 00:11:11.221 align:start position:0% switch after 00:11:11.221 --> 00:11:12.532 align:start position:0% switch after that. <00:11:11.645>On <00:11:12.069>Pro, <00:11:12.493>Max, 00:11:12.532 --> 00:11:12.568 align:start position:0% that. On Pro, Max, 00:11:12.568 --> 00:11:13.935 align:start position:0% that. On Pro, Max, Team, <00:11:12.780>or <00:11:12.992>Enterprise <00:11:13.204>plans, <00:11:13.416>fast 00:11:13.935 --> 00:11:13.971 align:start position:0% Team, or Enterprise plans, fast 00:11:13.971 --> 00:11:15.190 align:start position:0% Team, or Enterprise plans, fast mode <00:11:14.257>is <00:11:14.543>not <00:11:14.829>included <00:11:15.115>in 00:11:15.190 --> 00:11:15.227 align:start position:0% mode is not included in 00:11:15.227 --> 00:11:16.076 align:start position:0% mode is not included in your 00:11:16.076 --> 00:11:16.113 align:start position:0% your 00:11:16.113 --> 00:11:17.012 align:start position:0% your subscription <00:11:16.355>limits. <00:11:16.597>It <00:11:16.839>is 00:11:17.012 --> 00:11:17.041 align:start position:0% subscription limits. It is 00:11:17.041 --> 00:11:17.756 align:start position:0% subscription limits. It is paid <00:11:17.260>for <00:11:17.479>from <00:11:17.698>the 00:11:17.756 --> 00:11:17.785 align:start position:0% paid for from the 00:11:17.785 --> 00:11:18.672 align:start position:0% paid for from the additional 00:11:18.672 --> 00:11:18.700 align:start position:0% additional 00:11:18.700 --> 00:11:19.416 align:start position:0% additional usage <00:11:19.387>quota. 00:11:19.416 --> 00:11:19.516 align:start position:0% usage quota. 00:11:19.516 --> 00:11:20.670 align:start position:0% usage quota. Start <00:11:19.889>with <00:11:20.262>{slash} <00:11:20.635>status 00:11:20.670 --> 00:11:20.702 align:start position:0% Start with {slash} status 00:11:20.702 --> 00:11:21.889 align:start position:0% Start with {slash} status to <00:11:20.913>find <00:11:21.124>out <00:11:21.335>how <00:11:21.546>you <00:11:21.757>are 00:11:21.889 --> 00:11:21.922 align:start position:0% to find out how you are 00:11:21.922 --> 00:11:23.011 align:start position:0% to find out how you are logged <00:11:22.978>in. 00:11:23.011 --> 00:11:23.111 align:start position:0% logged in. 00:11:23.111 --> 00:11:24.174 align:start position:0% logged in. Pro <00:11:23.434>and <00:11:23.757>Max <00:11:24.080>users 00:11:24.174 --> 00:11:24.204 align:start position:0% Pro and Max users 00:11:24.204 --> 00:11:24.872 align:start position:0% Pro and Max users check 00:11:24.872 --> 00:11:24.903 align:start position:0% check 00:11:24.903 --> 00:11:26.087 align:start position:0% check the <00:11:25.196>settings <00:11:25.489>and <00:11:25.782>then 00:11:26.087 --> 00:11:26.118 align:start position:0% the settings and then 00:11:26.118 --> 00:11:27.302 align:start position:0% the settings and then usage <00:11:26.573>on <00:11:27.028>Claude.ai 00:11:27.302 --> 00:11:28.255 align:start position:0% usage on Claude.ai . <00:11:27.755>Team <00:11:28.208>and 00:11:28.255 --> 00:11:28.278 align:start position:0% . Team and 00:11:28.278 --> 00:11:28.872 align:start position:0% . Team and corporate 00:11:28.872 --> 00:11:28.894 align:start position:0% corporate 00:11:28.894 --> 00:11:29.260 align:start position:0% corporate account costs 00:11:29.260 --> 00:11:29.283 align:start position:0% account costs 00:11:29.283 --> 00:11:29.991 align:start position:0% account costs are <00:11:29.473>deducted <00:11:29.663>from <00:11:29.853>the 00:11:29.991 --> 00:11:30.014 align:start position:0% are deducted from the 00:11:30.014 --> 00:11:30.791 align:start position:0% are deducted from the organization's <00:11:30.768>quota. 00:11:30.791 --> 00:11:30.891 align:start position:0% organization's quota. 00:11:30.891 --> 00:11:31.852 align:start position:0% organization's quota. Users <00:11:31.359>enter <00:11:31.827>{ 00:11:31.852 --> 00:11:32.789 align:start position:0% Users enter { slash} <00:11:32.016>usage <00:11:32.180>to <00:11:32.344>view 00:11:32.789 --> 00:11:32.814 align:start position:0% slash} usage to view 00:11:32.814 --> 00:11:33.480 align:start position:0% slash} usage to view their <00:11:33.122>usage, <00:11:33.430>and 00:11:33.480 --> 00:11:33.504 align:start position:0% their usage, and 00:11:33.504 --> 00:11:34.195 align:start position:0% their usage, and administrators 00:11:34.195 --> 00:11:34.220 align:start position:0% administrators 00:11:34.220 --> 00:11:35.033 align:start position:0% administrators see <00:11:34.392>it <00:11:34.564>in <00:11:34.736>the 00:11:35.033 --> 00:11:35.058 align:start position:0% see it in the 00:11:35.058 --> 00:11:36.153 align:start position:0% see it in the control <00:11:35.286>panel. <00:11:35.514>API <00:11:35.742>clients 00:11:36.153 --> 00:11:36.182 align:start position:0% control panel. API clients 00:11:36.182 --> 00:11:36.885 align:start position:0% control panel. API clients in <00:11:36.357>the <00:11:36.532>Claude <00:11:36.707>console 00:11:36.885 --> 00:11:36.914 align:start position:0% in the Claude console 00:11:36.914 --> 00:11:37.645 align:start position:0% in the Claude console go <00:11:37.221>to <00:11:37.528>the 00:11:37.645 --> 00:11:37.675 align:start position:0% go to the 00:11:37.675 --> 00:11:38.553 align:start position:0% go to the costs <00:11:38.084>section <00:11:38.493>and 00:11:38.553 --> 00:11:38.582 align:start position:0% costs section and 00:11:38.582 --> 00:11:39.723 align:start position:0% costs section and select <00:11:39.196>"speed 00:11:39.723 --> 00:11:40.776 align:start position:0% select "speed " <00:11:39.979>from <00:11:40.235>the <00:11:40.491>grouping <00:11:40.747>menu, 00:11:40.776 --> 00:11:40.806 align:start position:0% " from the grouping menu, 00:11:40.806 --> 00:11:41.713 align:start position:0% " from the grouping menu, for <00:11:41.010>a 00:11:41.713 --> 00:11:41.742 align:start position:0% for a 00:11:41.742 --> 00:11:42.474 align:start position:0% for a preview. 00:11:42.474 --> 00:11:42.574 align:start position:0% preview. 00:11:42.574 --> 00:11:43.582 align:start position:0% preview. Free <00:11:43.168>evaluation 00:11:43.582 --> 00:11:43.608 align:start position:0% Free evaluation 00:11:43.608 --> 00:11:44.616 align:start position:0% Free evaluation accounts <00:11:44.461>cannot 00:11:44.616 --> 00:11:44.642 align:start position:0% accounts cannot 00:11:44.642 --> 00:11:45.573 align:start position:0% accounts cannot use <00:11:44.883>it <00:11:45.124>until <00:11:45.365>they 00:11:45.573 --> 00:11:45.599 align:start position:0% use it until they 00:11:45.599 --> 00:11:46.556 align:start position:0% use it until they top <00:11:45.909>up <00:11:46.219>their <00:11:46.529>balance. 00:11:46.556 --> 00:11:46.656 align:start position:0% top up their balance. 00:11:46.656 --> 00:11:47.861 align:start position:0% top up their balance. No <00:11:47.046>application <00:11:47.436>is <00:11:47.826>required. 00:11:47.861 --> 00:11:47.961 align:start position:0% No application is required. 00:11:47.961 --> 00:11:48.890 align:start position:0% No application is required. Just <00:11:48.184>add <00:11:48.407>the 00:11:48.890 --> 00:11:48.925 align:start position:0% Just add the 00:11:48.925 --> 00:11:50.233 align:start position:0% Just add the speed <00:11:49.544>parameter <00:11:50.163>" 00:11:50.233 --> 00:11:51.576 align:start position:0% speed parameter " fast" <00:11:50.465>to <00:11:50.697>your <00:11:50.929>API <00:11:51.161>request 00:11:51.576 --> 00:11:53.463 align:start position:0% fast" to your API request . <00:11:51.818>This <00:11:52.060>already <00:11:52.302>works <00:11:52.544>in <00:11:52.786>the <00:11:53.028>API <00:11:53.270>and 00:11:53.463 --> 00:11:53.511 align:start position:0% . This already works in the API and 00:11:53.511 --> 00:11:55.481 align:start position:0% . This already works in the API and Microsoft <00:11:53.956>Azure <00:11:54.401>Foundry. <00:11:54.846>Google <00:11:55.291>Cloud 00:11:55.481 --> 00:11:55.519 align:start position:0% Microsoft Azure Foundry. Google Cloud 00:11:55.519 --> 00:11:56.922 align:start position:0% Microsoft Azure Foundry. Google Cloud is <00:11:56.201>coming <00:11:56.883>soon, 00:11:56.922 --> 00:11:56.960 align:start position:0% is coming soon, 00:11:56.960 --> 00:11:58.022 align:start position:0% is coming soon, but <00:11:57.149>AWS <00:11:57.338>Bedrock <00:11:57.527>and <00:11:57.716>other 00:11:58.022 --> 00:11:58.060 align:start position:0% but AWS Bedrock and other 00:11:58.060 --> 00:11:59.274 align:start position:0% but AWS Bedrock and other platforms 00:11:59.274 --> 00:11:59.312 align:start position:0% platforms 00:11:59.312 --> 00:12:00.792 align:start position:0% platforms don't <00:11:59.792>support <00:12:00.272>it <00:12:00.752>yet. 00:12:00.792 --> 00:12:00.892 align:start position:0% don't support it yet. 00:12:00.892 --> 00:12:02.137 align:start position:0% don't support it yet. In <00:12:01.134>Claude <00:12:01.376>code, <00:12:01.618>you <00:12:01.860>type <00:12:02.102>{ 00:12:02.137 --> 00:12:03.244 align:start position:0% In Claude code, you type { slash} <00:12:02.437>fast, <00:12:02.737>space <00:12:03.037>to 00:12:03.244 --> 00:12:03.279 align:start position:0% slash} fast, space to 00:12:03.279 --> 00:12:04.525 align:start position:0% slash} fast, space to toggle, <00:12:03.798>Enter <00:12:04.317>to 00:12:04.525 --> 00:12:04.559 align:start position:0% toggle, Enter to 00:12:04.559 --> 00:12:05.597 align:start position:0% toggle, Enter to confirm, <00:12:05.528>and 00:12:05.597 --> 00:12:05.632 align:start position:0% confirm, and 00:12:05.632 --> 00:12:06.774 align:start position:0% confirm, and you <00:12:05.961>get <00:12:06.290>fast 00:12:06.774 --> 00:12:06.809 align:start position:0% you get fast 00:12:06.809 --> 00:12:07.708 align:start position:0% you get fast mode <00:12:07.189>along <00:12:07.569>with 00:12:07.708 --> 00:12:07.743 align:start position:0% mode along with 00:12:07.743 --> 00:12:08.885 align:start position:0% mode along with a <00:12:08.400>little 00:12:08.885 --> 00:12:08.919 align:start position:0% a little 00:12:08.919 --> 00:12:10.096 align:start position:0% a little lightning <00:12:09.126>bolt <00:12:09.333>icon <00:12:09.540>next <00:12:09.747>to <00:12:09.954>the 00:12:10.096 --> 00:12:10.130 align:start position:0% lightning bolt icon next to the 00:12:10.130 --> 00:12:11.658 align:start position:0% lightning bolt icon next to the prompt. 00:12:11.658 --> 00:12:11.693 align:start position:0% prompt. 00:12:11.693 --> 00:12:13.054 align:start position:0% prompt. VS <00:12:11.876>Code <00:12:12.059>users <00:12:12.242>can <00:12:12.425>enable 00:12:13.054 --> 00:12:13.089 align:start position:0% VS Code users can enable 00:12:13.089 --> 00:12:14.310 align:start position:0% VS Code users can enable it <00:12:13.280>with <00:12:13.471>a <00:12:13.662>single <00:12:13.853>click <00:12:14.044>in <00:12:14.235>the 00:12:14.310 --> 00:12:14.345 align:start position:0% it with a single click in the 00:12:14.345 --> 00:12:15.649 align:start position:0% it with a single click in the plugin. <00:12:15.187>You 00:12:15.649 --> 00:12:15.688 align:start position:0% plugin. You 00:12:15.688 --> 00:12:17.189 align:start position:0% plugin. You can <00:12:16.053>even <00:12:16.418>type <00:12:16.783>{slash} <00:12:17.148>fast 00:12:17.189 --> 00:12:17.228 align:start position:0% can even type {slash} fast 00:12:17.228 --> 00:12:17.998 align:start position:0% can even type {slash} fast while <00:12:17.433>Claude <00:12:17.638>is <00:12:17.843>still 00:12:17.998 --> 00:12:18.036 align:start position:0% while Claude is still 00:12:18.036 --> 00:12:19.133 align:start position:0% while Claude is still responding. <00:12:18.951>It 00:12:19.133 --> 00:12:19.164 align:start position:0% responding. It 00:12:19.164 --> 00:12:20.353 align:start position:0% responding. It will <00:12:19.430>complete <00:12:19.696>this <00:12:19.962>stage <00:12:20.228>at 00:12:20.353 --> 00:12:20.384 align:start position:0% will complete this stage at 00:12:20.384 --> 00:12:21.054 align:start position:0% will complete this stage at standard 00:12:21.054 --> 00:12:21.085 align:start position:0% standard 00:12:21.085 --> 00:12:21.756 align:start position:0% standard speed, <00:12:21.390>and <00:12:21.695>the 00:12:21.756 --> 00:12:21.786 align:start position:0% speed, and the 00:12:21.786 --> 00:12:22.884 align:start position:0% speed, and the next <00:12:22.091>one <00:12:22.396>will <00:12:22.701>begin 00:12:22.884 --> 00:12:22.914 align:start position:0% next one will begin 00:12:22.914 --> 00:12:24.475 align:start position:0% next one will begin quickly. <00:12:23.644>Starting <00:12:24.374>with 00:12:24.475 --> 00:12:24.526 align:start position:0% quickly. Starting with 00:12:24.526 --> 00:12:26.351 align:start position:0% quickly. Starting with version <00:12:25.084>2.1.280, <00:12:25.642>fast 00:12:26.351 --> 00:12:26.402 align:start position:0% version 2.1.280, fast 00:12:26.402 --> 00:12:28.177 align:start position:0% version 2.1.280, fast mode <00:12:26.960>uses 00:12:28.177 --> 00:12:28.228 align:start position:0% mode uses 00:12:28.228 --> 00:12:28.887 align:start position:0% mode uses Opus <00:12:28.456>5.5 <00:12:28.684>by 00:12:28.887 --> 00:12:28.937 align:start position:0% Opus 5.5 by 00:12:28.937 --> 00:12:30.522 align:start position:0% Opus 5.5 by default. <00:12:30.407>You 00:12:30.522 --> 00:12:30.551 align:start position:0% default. You 00:12:30.551 --> 00:12:31.650 align:start position:0% default. You can <00:12:30.898>also <00:12:31.245>hard- 00:12:31.650 --> 00:12:31.679 align:start position:0% can also hard- 00:12:31.679 --> 00:12:32.779 align:start position:0% can also hard- code <00:12:32.084>fast <00:12:32.489>mode 00:12:32.779 --> 00:12:32.807 align:start position:0% code fast mode 00:12:32.807 --> 00:12:33.675 align:start position:0% code fast mode in <00:12:33.270>your 00:12:33.675 --> 00:12:33.704 align:start position:0% in your 00:12:33.704 --> 00:12:34.370 align:start position:0% in your user <00:12:34.341>profile. 00:12:34.370 --> 00:12:34.470 align:start position:0% user profile. 00:12:34.470 --> 00:12:35.517 align:start position:0% user profile. Switch <00:12:34.980>to <00:12:35.490>another, 00:12:35.517 --> 00:12:35.544 align:start position:0% Switch to another, 00:12:35.544 --> 00:12:36.296 align:start position:0% Switch to another, unsupported 00:12:36.296 --> 00:12:36.323 align:start position:0% unsupported 00:12:36.323 --> 00:12:37.290 align:start position:0% unsupported model, <00:12:36.753>like <00:12:37.183>Sonnet, 00:12:37.290 --> 00:12:37.317 align:start position:0% model, like Sonnet, 00:12:37.317 --> 00:12:38.284 align:start position:0% model, like Sonnet, with <00:12:37.621>{slash} <00:12:37.925>model <00:12:38.229>and 00:12:38.284 --> 00:12:38.311 align:start position:0% with {slash} model and 00:12:38.311 --> 00:12:39.064 align:start position:0% with {slash} model and it <00:12:38.553>will <00:12:38.795>turn <00:12:39.037>off. 00:12:39.064 --> 00:12:39.164 align:start position:0% it will turn off. 00:12:39.164 --> 00:12:40.192 align:start position:0% it will turn off. Switch <00:12:39.898>back 00:12:40.192 --> 00:12:40.221 align:start position:0% Switch back 00:12:40.221 --> 00:12:40.750 align:start position:0% Switch back to <00:12:40.358>Opus <00:12:40.495>5.5 <00:12:40.632>with 00:12:40.750 --> 00:12:40.779 align:start position:0% to Opus 5.5 with 00:12:40.779 --> 00:12:41.425 align:start position:0% to Opus 5.5 with 00:12:41.425 --> 00:12:41.455 align:start position:0% 00:12:41.455 --> 00:12:42.394 align:start position:0% the <00:12:41.675>settings <00:12:41.895>saved <00:12:42.115>and <00:12:42.335>the 00:12:42.394 --> 00:12:42.424 align:start position:0% the settings saved and the 00:12:42.424 --> 00:12:43.099 align:start position:0% the settings saved and the icon <00:12:42.806>will 00:12:43.099 --> 00:12:43.129 align:start position:0% icon will 00:12:43.129 --> 00:12:44.477 align:start position:0% icon will light <00:12:43.336>up <00:12:43.543>again. <00:12:43.750>In <00:12:43.957>cloud 00:12:44.477 --> 00:12:44.514 align:start position:0% light up again. In cloud 00:12:44.514 --> 00:12:45.813 align:start position:0% light up again. In cloud sessions, <00:12:45.145>the <00:12:45.776>{slash} 00:12:45.813 --> 00:12:45.850 align:start position:0% sessions, the {slash} 00:12:45.850 --> 00:12:46.889 align:start position:0% sessions, the {slash} fast <00:12:45.998>on <00:12:46.146>command <00:12:46.294>only <00:12:46.442>applies <00:12:46.590>to <00:12:46.738>the 00:12:46.889 --> 00:12:46.926 align:start position:0% fast on command only applies to the 00:12:46.926 --> 00:12:47.892 align:start position:0% fast on command only applies to the current <00:12:47.854>session. 00:12:47.892 --> 00:12:47.992 align:start position:0% current session. 00:12:47.992 --> 00:12:49.096 align:start position:0% current session. The <00:12:48.204>main <00:12:48.416>concern 00:12:49.096 --> 00:12:49.124 align:start position:0% The main concern 00:12:49.124 --> 00:12:50.143 align:start position:0% The main concern in <00:12:49.364>the <00:12:49.604>comments <00:12:49.844>was: <00:12:50.084>“ 00:12:50.143 --> 00:12:51.191 align:start position:0% in the comments was: “ If <00:12:50.376>this <00:12:50.609>is <00:12:50.842>enabled <00:12:51.075>by 00:12:51.191 --> 00:12:51.219 align:start position:0% If this is enabled by 00:12:51.219 --> 00:12:51.984 align:start position:0% If this is enabled by default, 00:12:51.984 --> 00:12:52.012 align:start position:0% default, 00:12:52.012 --> 00:12:53.003 align:start position:0% default, half <00:12:52.266>of <00:12:52.520>us <00:12:52.774>could 00:12:53.003 --> 00:12:53.031 align:start position:0% half of us could 00:12:53.031 --> 00:12:54.107 align:start position:0% half of us could burn <00:12:53.512>$200 <00:12:53.993>in 00:12:54.107 --> 00:12:54.136 align:start position:0% burn $200 in 00:12:54.136 --> 00:12:55.013 align:start position:0% burn $200 in tokens <00:12:54.371>before <00:12:54.606>we <00:12:54.841>even 00:12:55.013 --> 00:12:55.042 align:start position:0% tokens before we even 00:12:55.042 --> 00:12:55.928 align:start position:0% tokens before we even notice.” <00:12:55.680>Therefore, 00:12:55.928 --> 00:12:55.959 align:start position:0% notice.” Therefore, 00:12:55.959 --> 00:12:56.827 align:start position:0% notice.” Therefore, administrators 00:12:56.827 --> 00:12:56.858 align:start position:0% administrators 00:12:56.858 --> 00:12:57.974 align:start position:0% administrators can <00:12:57.385>set <00:12:57.912>an 00:12:57.974 --> 00:12:58.966 align:start position:0% can set an opt-in <00:12:58.222>for <00:12:58.470>fast 00:12:58.966 --> 00:12:58.997 align:start position:0% opt-in for fast 00:12:58.997 --> 00:13:00.114 align:start position:0% opt-in for fast mode <00:12:59.307>in <00:12:59.617>settings.json <00:12:59.927>or 00:13:00.114 --> 00:13:00.145 align:start position:0% mode in settings.json or 00:13:00.145 --> 00:13:01.199 align:start position:0% mode in settings.json or at <00:13:00.393>the <00:13:00.641>server <00:13:00.889>level, <00:13:01.137>and 00:13:01.199 --> 00:13:01.230 align:start position:0% at the server level, and 00:13:01.230 --> 00:13:02.160 align:start position:0% at the server level, and each <00:13:01.540>new <00:13:01.850>session 00:13:02.160 --> 00:13:02.191 align:start position:0% each new session 00:13:02.191 --> 00:13:03.153 align:start position:0% each new session will <00:13:02.641>start <00:13:03.091>with 00:13:03.153 --> 00:13:03.184 align:start position:0% will start with 00:13:03.184 --> 00:13:04.207 align:start position:0% will start with fast 00:13:04.207 --> 00:13:04.238 align:start position:0% fast 00:13:04.238 --> 00:13:05.323 align:start position:0% fast mode <00:13:04.496>disabled <00:13:04.754>until <00:13:05.012>someone 00:13:05.323 --> 00:13:05.354 align:start position:0% mode disabled until someone 00:13:05.354 --> 00:13:06.468 align:start position:0% mode disabled until someone types <00:13:05.604>{slash} <00:13:05.854>fast. <00:13:06.104>This <00:13:06.354>is 00:13:06.468 --> 00:13:06.497 align:start position:0% types {slash} fast. This is 00:13:06.497 --> 00:13:07.415 align:start position:0% types {slash} fast. This is created <00:13:06.841>for <00:13:07.185>that 00:13:07.415 --> 00:13:07.443 align:start position:0% created for that 00:13:07.443 --> 00:13:08.562 align:start position:0% created for that colleague <00:13:07.773>who <00:13:08.103>runs 00:13:08.562 --> 00:13:08.590 align:start position:0% colleague who runs 00:13:08.590 --> 00:13:09.365 align:start position:0% colleague who runs more <00:13:08.714>than <00:13:08.838>a <00:13:08.962>dozen 00:13:09.365 --> 00:13:09.393 align:start position:0% more than a dozen 00:13:09.393 --> 00:13:10.512 align:start position:0% more than a dozen terminals <00:13:09.694>at <00:13:09.995>once 00:13:10.512 --> 00:13:11.224 align:start position:0% terminals at once . <00:13:10.722>Anthropic <00:13:10.932>also 00:13:11.224 --> 00:13:11.253 align:start position:0% . Anthropic also 00:13:11.253 --> 00:13:11.778 align:start position:0% . Anthropic also recently 00:13:11.778 --> 00:13:11.807 align:start position:0% recently 00:13:11.807 --> 00:13:12.536 align:start position:0% recently introduced <00:13:12.478>“ 00:13:12.536 --> 00:13:13.469 align:start position:0% introduced “ downtime <00:13:13.294>compression, 00:13:13.469 --> 00:13:13.498 align:start position:0% downtime compression, 00:13:13.498 --> 00:13:14.140 align:start position:0% downtime compression, ” <00:13:14.023>which 00:13:14.140 --> 00:13:14.169 align:start position:0% ” which 00:13:14.169 --> 00:13:15.219 align:start position:0% ” which fixes <00:13:14.577>one <00:13:14.985>very 00:13:15.219 --> 00:13:15.248 align:start position:0% fixes one very 00:13:15.248 --> 00:13:16.366 align:start position:0% fixes one very annoying <00:13:15.735>problem. <00:13:16.222>You 00:13:16.366 --> 00:13:16.402 align:start position:0% annoying problem. You 00:13:16.402 --> 00:13:17.333 align:start position:0% annoying problem. You spent <00:13:17.225>200,000 00:13:17.333 --> 00:13:17.368 align:start position:0% spent 200,000 00:13:17.368 --> 00:13:18.407 align:start position:0% spent 200,000 tokens <00:13:17.702>in <00:13:18.036>a <00:13:18.370>session, 00:13:18.407 --> 00:13:18.443 align:start position:0% tokens in a session, 00:13:18.443 --> 00:13:19.803 align:start position:0% tokens in a session, went <00:13:18.729>to <00:13:19.015>a <00:13:19.301>meeting, <00:13:19.587>the 00:13:19.803 --> 00:13:19.839 align:start position:0% went to a meeting, the 00:13:19.839 --> 00:13:21.200 align:start position:0% went to a meeting, the prompt <00:13:20.280>cache <00:13:20.721>ran <00:13:21.162>out, 00:13:21.200 --> 00:13:21.236 align:start position:0% prompt cache ran out, 00:13:21.236 --> 00:13:22.346 align:start position:0% prompt cache ran out, and <00:13:21.438>when <00:13:21.640>you <00:13:21.842>request 00:13:22.346 --> 00:13:22.382 align:start position:0% and when you request 00:13:22.382 --> 00:13:23.420 align:start position:0% and when you request one <00:13:22.716>line <00:13:23.050>of <00:13:23.384>code, 00:13:23.420 --> 00:13:23.456 align:start position:0% one line of code, 00:13:23.456 --> 00:13:24.709 align:start position:0% one line of code, the <00:13:23.635>system <00:13:23.814>has <00:13:23.993>to 00:13:24.709 --> 00:13:24.745 align:start position:0% the system has to 00:13:24.745 --> 00:13:26.034 align:start position:0% the system has to read <00:13:25.819>all 00:13:26.034 --> 00:13:26.070 align:start position:0% read all 00:13:26.070 --> 00:13:27.466 align:start position:0% read all 200,000 <00:13:26.249>again, <00:13:26.428>and <00:13:26.607>a <00:13:26.786>few <00:13:26.965>dollars 00:13:27.466 --> 00:13:27.502 align:start position:0% 200,000 again, and a few dollars 00:13:27.502 --> 00:13:28.577 align:start position:0% 200,000 again, and a few dollars just <00:13:28.541>disappear. 00:13:28.577 --> 00:13:29.425 align:start position:0% just disappear. 00:13:29.425 --> 00:13:30.560 align:start position:0% just disappear. Now, <00:13:30.426>before 00:13:30.560 --> 00:13:30.593 align:start position:0% Now, before 00:13:30.593 --> 00:13:31.828 align:start position:0% Now, before the <00:13:30.851>cache <00:13:31.109>runs <00:13:31.367>out, <00:13:31.625>Claude 00:13:31.828 --> 00:13:31.862 align:start position:0% the cache runs out, Claude 00:13:31.862 --> 00:13:32.930 align:start position:0% the cache runs out, Claude compresses <00:13:32.112>the <00:13:32.362>conversation <00:13:32.612>into <00:13:32.862>a 00:13:32.930 --> 00:13:32.963 align:start position:0% compresses the conversation into a 00:13:32.963 --> 00:13:34.165 align:start position:0% compresses the conversation into a short <00:13:33.531>summary <00:13:34.099>and 00:13:34.165 --> 00:13:34.199 align:start position:0% short summary and 00:13:34.199 --> 00:13:35.334 align:start position:0% short summary and continues <00:13:34.555>to <00:13:34.911>work <00:13:35.267>from 00:13:35.334 --> 00:13:35.367 align:start position:0% continues to work from 00:13:35.367 --> 00:13:37.278 align:start position:0% continues to work from it. <00:13:35.743>This <00:13:36.119>is <00:13:36.495>triggered 00:13:37.278 --> 00:13:37.811 align:start position:0% it. This is triggered when <00:13:37.669>you 00:13:37.811 --> 00:13:37.846 align:start position:0% when you 00:13:37.846 --> 00:13:38.841 align:start position:0% when you are using the 00:13:38.841 --> 00:13:38.877 align:start position:0% are using the 00:13:38.877 --> 00:13:39.694 align:start position:0% are using the hourly 00:13:39.694 --> 00:13:39.729 align:start position:0% hourly 00:13:39.729 --> 00:13:40.582 align:start position:0% hourly prompt <00:13:40.049>cache, <00:13:40.369>your 00:13:40.582 --> 00:13:40.618 align:start position:0% prompt cache, your 00:13:40.618 --> 00:13:42.003 align:start position:0% prompt cache, your context <00:13:41.257>reaches <00:13:41.896>200,000 00:13:42.003 --> 00:13:42.039 align:start position:0% context reaches 200,000 00:13:42.039 --> 00:13:43.389 align:start position:0% context reaches 200,000 tokens, <00:13:42.441>the <00:13:42.843>computer <00:13:43.245>is 00:13:43.389 --> 00:13:43.425 align:start position:0% tokens, the computer is 00:13:43.425 --> 00:13:44.419 align:start position:0% tokens, the computer is awake, <00:13:43.602>and <00:13:43.779>you <00:13:43.956>have <00:13:44.133>been 00:13:44.419 --> 00:13:44.455 align:start position:0% awake, and you have been 00:13:44.455 --> 00:13:45.734 align:start position:0% awake, and you have been inactive <00:13:44.857>for <00:13:45.259>about <00:13:45.661>54 00:13:45.734 --> 00:13:45.770 align:start position:0% inactive for about 54 00:13:45.770 --> 00:13:47.397 align:start position:0% inactive for about 54 minutes. <00:13:46.166>What <00:13:46.562>is <00:13:46.958>the <00:13:47.354>downside? 00:13:47.397 --> 00:13:47.555 align:start position:0% minutes. What is the downside? 00:13:47.555 --> 00:13:48.881 align:start position:0% minutes. What is the downside? One <00:13:47.929>developer <00:13:48.303>on <00:13:48.677>GitHub 00:13:48.881 --> 00:13:48.915 align:start position:0% One developer on GitHub 00:13:48.915 --> 00:13:49.969 align:start position:0% One developer on GitHub reported <00:13:49.340>that <00:13:49.765>seven 00:13:49.969 --> 00:13:50.003 align:start position:0% reported that seven 00:13:50.003 --> 00:13:51.092 align:start position:0% reported that seven long <00:13:50.411>sessions <00:13:50.819>were 00:13:51.092 --> 00:13:51.126 align:start position:0% long sessions were 00:13:51.126 --> 00:13:52.384 align:start position:0% long sessions were compressed <00:13:51.406>overnight, <00:13:51.686>and <00:13:51.966>the <00:13:52.246>AI 00:13:52.384 --> 00:13:52.418 align:start position:0% compressed overnight, and the AI 00:13:52.418 --> 00:13:53.439 align:start position:0% compressed overnight, and the AI forgot <00:13:52.826>details <00:13:53.234>about 00:13:53.439 --> 00:13:53.473 align:start position:0% forgot details about 00:13:53.473 --> 00:13:54.459 align:start position:0% forgot details about basic <00:13:53.915>technical 00:13:54.459 --> 00:13:54.493 align:start position:0% basic technical 00:13:54.493 --> 00:13:56.907 align:start position:0% basic technical decisions. <00:13:55.426>Anthropic <00:13:56.359>quickly 00:13:56.907 --> 00:13:56.953 align:start position:0% decisions. Anthropic quickly 00:13:56.953 --> 00:13:58.639 align:start position:0% decisions. Anthropic quickly added <00:13:57.454>a <00:13:57.955>switch: <00:13:58.456>idle 00:13:58.639 --> 00:13:58.685 align:start position:0% added a switch: idle 00:13:58.685 --> 00:14:00.189 align:start position:0% added a switch: idle compaction <00:13:58.989>false <00:13:59.293>in <00:13:59.597>settings.json 00:14:00.189 --> 00:14:00.235 align:start position:0% compaction false in settings.json 00:14:00.235 --> 00:14:01.101 align:start position:0% compaction false in settings.json or <00:14:00.554>an 00:14:01.101 --> 00:14:01.146 align:start position:0% or an 00:14:01.146 --> 00:14:02.423 align:start position:0% or an environment <00:14:01.647>variable <00:14:02.148>to 00:14:02.423 --> 00:14:02.468 align:start position:0% environment variable to 00:14:02.468 --> 00:14:03.836 align:start position:0% environment variable to change <00:14:02.764>the <00:14:03.060>threshold <00:14:03.356>from <00:14:03.652>the 00:14:03.836 --> 00:14:03.881 align:start position:0% change the threshold from the 00:14:03.881 --> 00:14:05.295 align:start position:0% change the threshold from the default <00:14:05.249>200,000. 00:14:05.295 --> 00:14:06.317 align:start position:0% default 200,000. 00:14:06.317 --> 00:14:07.828 align:start position:0% Okay, <00:14:06.509>one <00:14:06.701>last <00:14:06.893>thing, <00:14:07.085>and <00:14:07.277>this <00:14:07.469>is <00:14:07.661>a 00:14:07.828 --> 00:14:07.869 align:start position:0% Okay, one last thing, and this is a 00:14:07.869 --> 00:14:09.796 align:start position:0% Okay, one last thing, and this is a surprise. <00:14:08.401>Meta <00:14:08.933>has <00:14:09.465>just 00:14:09.796 --> 00:14:09.829 align:start position:0% surprise. Meta has just 00:14:09.829 --> 00:14:10.986 align:start position:0% surprise. Meta has just released <00:14:10.203>Muse <00:14:10.577>Spark <00:14:10.951>Open, 00:14:10.986 --> 00:14:11.019 align:start position:0% released Muse Spark Open, 00:14:11.019 --> 00:14:11.714 align:start position:0% released Muse Spark Open, calling <00:14:11.300>it <00:14:11.581>the 00:14:11.714 --> 00:14:11.747 align:start position:0% calling it the 00:14:11.747 --> 00:14:12.606 align:start position:0% calling it the most powerful 00:14:12.606 --> 00:14:13.928 align:start position:0% most powerful 00:14:13.928 --> 00:14:13.962 align:start position:0% 00:14:13.962 --> 00:14:15.218 align:start position:0% open- 00:14:15.218 --> 00:14:15.251 align:start position:0% open- 00:14:15.251 --> 00:14:16.145 align:start position:0% open- source <00:14:15.416>American <00:14:15.581>language <00:14:15.746>model. <00:14:15.911>It <00:14:16.076>is 00:14:16.145 --> 00:14:16.178 align:start position:0% source American language model. It is 00:14:16.178 --> 00:14:16.642 align:start position:0% source American language model. It is natively 00:14:16.642 --> 00:14:16.675 align:start position:0% natively 00:14:16.675 --> 00:14:17.934 align:start position:0% natively multimodal, <00:14:17.735>has 00:14:17.934 --> 00:14:17.967 align:start position:0% multimodal, has 00:14:17.967 --> 00:14:18.696 align:start position:0% multimodal, has 816 <00:14:18.100>billion 00:14:18.696 --> 00:14:18.729 align:start position:0% 816 billion 00:14:18.729 --> 00:14:20.021 align:start position:0% 816 billion total <00:14:19.359>parameters 00:14:20.021 --> 00:14:21.314 align:start position:0% total parameters , <00:14:20.272>42 <00:14:20.523>billion <00:14:20.774>active <00:14:21.025>ones, <00:14:21.276>a 00:14:21.314 --> 00:14:21.347 align:start position:0% , 42 billion active ones, a 00:14:21.347 --> 00:14:22.539 align:start position:0% , 42 billion active ones, a context <00:14:21.590>of <00:14:21.833>1 <00:14:22.076>million 00:14:22.539 --> 00:14:22.573 align:start position:0% context of 1 million 00:14:22.573 --> 00:14:23.865 align:start position:0% context of 1 million tokens, <00:14:22.846>and <00:14:23.119>is <00:14:23.392>built <00:14:23.665>for 00:14:23.865 --> 00:14:23.898 align:start position:0% tokens, and is built for 00:14:23.898 --> 00:14:24.792 align:start position:0% tokens, and is built for programming, 00:14:24.792 --> 00:14:24.825 align:start position:0% programming, 00:14:24.825 --> 00:14:25.753 align:start position:0% programming, agent <00:14:25.720>work, 00:14:25.753 --> 00:14:25.786 align:start position:0% agent work, 00:14:25.786 --> 00:14:26.913 align:start position:0% agent work, long-running <00:14:26.316>tasks, <00:14:26.846>and 00:14:26.913 --> 00:14:26.946 align:start position:0% long-running tasks, and 00:14:26.946 --> 00:14:27.476 align:start position:0% long-running tasks, and complex 00:14:27.476 --> 00:14:27.509 align:start position:0% complex 00:14:27.509 --> 00:14:28.437 align:start position:0% complex multi-step 00:14:28.437 --> 00:14:28.470 align:start position:0% multi-step 00:14:28.470 --> 00:14:29.568 align:start position:0% multi-step reasoning. <00:14:28.666>The <00:14:28.862>idea <00:14:29.058>behind <00:14:29.254>Meta <00:14:29.450>is 00:14:29.568 --> 00:14:29.596 align:start position:0% reasoning. The idea behind Meta is 00:14:29.596 --> 00:14:30.503 align:start position:0% reasoning. The idea behind Meta is that 00:14:30.503 --> 00:14:30.531 align:start position:0% that 00:14:30.531 --> 00:14:31.324 align:start position:0% that everyone <00:14:30.701>has <00:14:30.871>the <00:14:31.041>right 00:14:31.324 --> 00:14:31.353 align:start position:0% everyone has the right 00:14:31.353 --> 00:14:32.401 align:start position:0% everyone has the right to <00:14:31.622>run <00:14:31.891>advanced 00:14:32.401 --> 00:14:32.429 align:start position:0% to run advanced 00:14:32.429 --> 00:14:33.336 align:start position:0% to run advanced models <00:14:32.599>on <00:14:32.769>their <00:14:32.939>own 00:14:33.336 --> 00:14:33.364 align:start position:0% models on their own 00:14:33.364 --> 00:14:34.299 align:start position:0% models on their own devices <00:14:33.656>at <00:14:33.948>home, <00:14:34.240>and 00:14:34.299 --> 00:14:34.327 align:start position:0% devices at home, and 00:14:34.327 --> 00:14:35.092 align:start position:0% devices at home, and Mark <00:14:34.582>Zuckerberg 00:14:35.092 --> 00:14:35.121 align:start position:0% Mark Zuckerberg 00:14:35.121 --> 00:14:35.886 align:start position:0% Mark Zuckerberg has <00:14:35.255>said <00:14:35.389>that <00:14:35.523>it <00:14:35.657>is 00:14:35.886 --> 00:14:35.914 align:start position:0% has said that it is 00:14:35.914 --> 00:14:36.481 align:start position:0% has said that it is specifically 00:14:36.481 --> 00:14:36.509 align:start position:0% specifically 00:14:36.509 --> 00:14:37.585 align:start position:0% specifically optimized <00:14:36.764>for <00:14:37.019>the <00:14:37.274>Mac <00:14:37.529>in 00:14:37.585 --> 00:14:37.614 align:start position:0% optimized for the Mac in 00:14:37.614 --> 00:14:38.520 align:start position:0% optimized for the Mac in collaboration <00:14:38.237>with 00:14:38.520 --> 00:14:38.549 align:start position:0% collaboration with 00:14:38.549 --> 00:14:39.455 align:start position:0% collaboration with Apple's <00:14:38.931>new <00:14:39.313>CEO 00:14:39.455 --> 00:14:39.484 align:start position:0% Apple's new CEO 00:14:39.484 --> 00:14:40.746 align:start position:0% Apple's new CEO John <00:14:39.803>Turnus. <00:14:40.122>It <00:14:40.441>is 00:14:40.746 --> 00:14:40.784 align:start position:0% John Turnus. It is 00:14:40.784 --> 00:14:42.038 align:start position:0% John Turnus. It is similar <00:14:41.019>to <00:14:41.254>Muse <00:14:41.489>Spark <00:14:41.724>1.3 <00:14:41.959>and 00:14:42.038 --> 00:14:42.076 align:start position:0% similar to Muse Spark 1.3 and 00:14:42.076 --> 00:14:43.558 align:start position:0% similar to Muse Spark 1.3 and uses <00:14:42.551>the <00:14:43.026>latest 00:14:43.558 --> 00:14:43.596 align:start position:0% uses the latest 00:14:43.596 --> 00:14:44.963 align:start position:0% uses the latest Apple <00:14:43.975>MLX <00:14:44.354>framework <00:14:44.733>to 00:14:44.963 --> 00:14:45.001 align:start position:0% Apple MLX framework to 00:14:45.001 --> 00:14:45.837 align:start position:0% Apple MLX framework to 00:14:45.837 --> 00:14:45.875 align:start position:0% 00:14:45.875 --> 00:14:47.015 align:start position:0% make the most of 00:14:47.015 --> 00:14:48.496 align:start position:0% make the most of 00:14:48.496 --> 00:14:48.534 align:start position:0% 00:14:48.534 --> 00:14:49.902 align:start position:0% Apple <00:14:48.686>silicon's <00:14:48.838>unified <00:14:48.990>memory, <00:14:49.142>so <00:14:49.294>it <00:14:49.446>runs 00:14:49.902 --> 00:14:49.940 align:start position:0% Apple silicon's unified memory, so it runs 00:14:49.940 --> 00:14:51.422 align:start position:0% Apple silicon's unified memory, so it runs locally <00:14:50.263>on <00:14:50.586>two <00:14:50.909>M5 <00:14:51.232>Ultra 00:14:51.422 --> 00:14:51.460 align:start position:0% locally on two M5 Ultra 00:14:51.460 --> 00:14:52.944 align:start position:0% locally on two M5 Ultra Mac <00:14:51.796>Studios. <00:14:52.132>The 00:14:52.944 --> 00:14:52.978 align:start position:0% Mac Studios. The 00:14:52.978 --> 00:14:53.690 align:start position:0% Mac Studios. The M5 <00:14:53.088>Ultra's <00:14:53.198>integrated <00:14:53.308>GPU <00:14:53.418>can 00:14:53.690 --> 00:14:53.724 align:start position:0% M5 Ultra's integrated GPU can 00:14:53.724 --> 00:14:55.045 align:start position:0% M5 Ultra's integrated GPU can use <00:14:54.333>up <00:14:54.942>to 00:14:55.045 --> 00:14:55.078 align:start position:0% use up to 00:14:55.078 --> 00:14:56.399 align:start position:0% use up to 512GB <00:14:55.400>of <00:14:55.722>memory <00:14:56.044>per <00:14:56.366>machine, 00:14:56.399 --> 00:14:56.433 align:start position:0% 512GB of memory per machine, 00:14:56.433 --> 00:14:57.517 align:start position:0% 512GB of memory per machine, so <00:14:56.704>two <00:14:56.975>devices 00:14:57.517 --> 00:14:57.551 align:start position:0% so two devices 00:14:57.551 --> 00:14:58.804 align:start position:0% so two devices can <00:14:57.856>accommodate <00:14:58.161>the <00:14:58.466>entire <00:14:58.771>model, 00:14:58.804 --> 00:14:58.838 align:start position:0% can accommodate the entire model, 00:14:58.838 --> 00:14:59.550 align:start position:0% can accommodate the entire model, and <00:14:58.940>the 00:14:59.550 --> 00:15:00.871 align:start position:0% and the 00:15:00.871 --> 00:15:00.905 align:start position:0% 00:15:00.905 --> 00:15:01.955 align:start position:0% 1.2TB/s <00:15:01.074>memory <00:15:01.243>bandwidth <00:15:01.412>is <00:15:01.581>sufficient <00:15:01.750>to 00:15:01.955 --> 00:15:01.989 align:start position:0% 1.2TB/s memory bandwidth is sufficient to 00:15:01.989 --> 00:15:03.039 align:start position:0% 1.2TB/s memory bandwidth is sufficient to read <00:15:02.344>the <00:15:02.699>small 00:15:03.039 --> 00:15:03.072 align:start position:0% read the small 00:15:03.072 --> 00:15:04.326 align:start position:0% read the small memory <00:15:04.292>chunks 00:15:04.326 --> 00:15:04.360 align:start position:0% memory chunks 00:15:04.360 --> 00:15:05.274 align:start position:0% memory chunks needed <00:15:05.071>for 00:15:05.274 --> 00:15:05.308 align:start position:0% needed for 00:15:05.308 --> 00:15:06.632 align:start position:0% needed for each <00:15:05.955>token. <00:15:06.602>So 00:15:06.632 --> 00:15:06.663 align:start position:0% each token. So 00:15:06.663 --> 00:15:07.460 align:start position:0% each token. So this <00:15:06.877>pair <00:15:07.091>works 00:15:07.460 --> 00:15:07.490 align:start position:0% this pair works 00:15:07.490 --> 00:15:08.532 align:start position:0% this pair works about <00:15:08.287>as 00:15:08.532 --> 00:15:08.563 align:start position:0% about as 00:15:08.563 --> 00:15:09.115 align:start position:0% about as fast <00:15:08.777>as <00:15:08.991>a 00:15:09.115 --> 00:15:09.145 align:start position:0% fast as a 00:15:09.145 --> 00:15:10.340 align:start position:0% fast as a supercomputer <00:15:09.413>with <00:15:09.681>1 <00:15:09.949>TB <00:15:10.217>of 00:15:10.340 --> 00:15:10.371 align:start position:0% supercomputer with 1 TB of 00:15:10.371 --> 00:15:11.107 align:start position:0% supercomputer with 1 TB of unified 00:15:11.107 --> 00:15:11.137 align:start position:0% unified 00:15:11.137 --> 00:15:12.540 align:start position:0% unified memory. <00:15:11.475>Although <00:15:11.813>“at <00:15:12.151>home 00:15:12.540 --> 00:15:13.972 align:start position:0% memory. Although “at home ” <00:15:13.004>is <00:15:13.468>an <00:15:13.932>exaggeration 00:15:13.972 --> 00:15:14.010 align:start position:0% ” is an exaggeration 00:15:14.010 --> 00:15:15.365 align:start position:0% ” is an exaggeration when <00:15:14.319>two <00:15:14.628>512GB <00:15:14.937>Mac <00:15:15.246>Studios 00:15:15.365 --> 00:15:15.404 align:start position:0% when two 512GB Mac Studios 00:15:15.404 --> 00:15:16.913 align:start position:0% when two 512GB Mac Studios cost <00:15:16.797>over 00:15:16.913 --> 00:15:16.951 align:start position:0% cost over 00:15:16.951 --> 00:15:18.306 align:start position:0% cost over $40,000, <00:15:17.206>and <00:15:17.461>Apple <00:15:17.716>and <00:15:17.971>Meta <00:15:18.226>AI 00:15:18.306 --> 00:15:18.344 align:start position:0% $40,000, and Apple and Meta AI 00:15:18.344 --> 00:15:19.389 align:start position:0% $40,000, and Apple and Meta AI have <00:15:18.712>historically <00:15:19.080>been 00:15:19.389 --> 00:15:19.428 align:start position:0% have historically been 00:15:19.428 --> 00:15:20.821 align:start position:0% have historically been pretty <00:15:20.511>cold 00:15:20.821 --> 00:15:20.859 align:start position:0% pretty cold 00:15:20.859 --> 00:15:22.059 align:start position:0% pretty cold towards <00:15:21.194>each <00:15:21.529>other, <00:15:21.864>Axios 00:15:22.059 --> 00:15:22.098 align:start position:0% towards each other, Axios 00:15:22.098 --> 00:15:23.491 align:start position:0% towards each other, Axios called <00:15:22.252>it <00:15:22.406>“the <00:15:22.560>burying <00:15:22.714>of <00:15:22.868>the 00:15:23.491 --> 00:15:23.529 align:start position:0% called it “the burying of the 00:15:23.529 --> 00:15:24.536 align:start position:0% called it “the burying of the hatchet.” 00:15:24.536 --> 00:15:24.779 align:start position:0% hatchet.” 00:15:24.779 --> 00:15:25.735 align:start position:0% hatchet.” Alexander <00:15:25.701>Wang, 00:15:25.735 --> 00:15:25.769 align:start position:0% Alexander Wang, 00:15:25.769 --> 00:15:26.827 align:start position:0% Alexander Wang, CEO <00:15:26.110>of <00:15:26.451>Meta <00:15:26.792>AI, 00:15:26.827 --> 00:15:26.862 align:start position:0% CEO of Meta AI, 00:15:26.862 --> 00:15:28.091 align:start position:0% CEO of Meta AI, said <00:15:27.180>they <00:15:27.498>were <00:15:27.816>happy 00:15:28.091 --> 00:15:28.125 align:start position:0% said they were happy 00:15:28.125 --> 00:15:29.252 align:start position:0% said they were happy to <00:15:28.213>see <00:15:28.301>the <00:15:28.389>reconciliation <00:15:28.477>between <00:15:28.565>the 00:15:29.252 --> 00:15:29.286 align:start position:0% to see the reconciliation between the 00:15:29.286 --> 00:15:30.345 align:start position:0% to see the reconciliation between the companies, <00:15:29.548>which <00:15:29.810>they <00:15:30.072>have 00:15:30.345 --> 00:15:30.379 align:start position:0% companies, which they have 00:15:30.379 --> 00:15:31.642 align:start position:0% companies, which they have been <00:15:30.663>collaborating <00:15:30.947>for <00:15:31.231>several 00:15:31.642 --> 00:15:31.677 align:start position:0% been collaborating for several 00:15:31.677 --> 00:15:32.735 align:start position:0% been collaborating for several months, <00:15:32.001>and <00:15:32.325>specifically 00:15:32.735 --> 00:15:32.769 align:start position:0% months, and specifically 00:15:32.769 --> 00:15:34.101 align:start position:0% months, and specifically thanked <00:15:33.418>former 00:15:34.101 --> 00:15:34.135 align:start position:0% thanked former 00:15:34.135 --> 00:15:35.160 align:start position:0% thanked former Apple <00:15:34.989>executive 00:15:35.160 --> 00:15:35.194 align:start position:0% Apple executive 00:15:35.194 --> 00:15:36.287 align:start position:0% Apple executive Daniel <00:15:35.512>Gross <00:15:35.830>for <00:15:36.148>his 00:15:36.287 --> 00:15:36.321 align:start position:0% Daniel Gross for his 00:15:36.321 --> 00:15:39.531 align:start position:0% Daniel Gross for his help <00:15:37.321>in <00:15:38.321>this. <00:15:39.321>Meta 00:15:39.531 --> 00:15:39.584 align:start position:0% help in this. Meta 00:15:39.584 --> 00:15:41.483 align:start position:0% help in this. Meta has <00:15:40.146>also <00:15:40.708>adapted <00:15:41.270>it 00:15:41.483 --> 00:15:41.535 align:start position:0% has also adapted it 00:15:41.535 --> 00:15:43.328 align:start position:0% has also adapted it for <00:15:41.883>SG <00:15:42.231>Lang <00:15:42.579>VLLM, <00:15:42.927>TensorRT <00:15:43.275>LLM, 00:15:43.328 --> 00:15:43.381 align:start position:0% for SG Lang VLLM, TensorRT LLM, 00:15:43.381 --> 00:15:45.332 align:start position:0% for SG Lang VLLM, TensorRT LLM, Hugging <00:15:43.855>Face <00:15:44.329>Transformers, <00:15:44.803>Apple <00:15:45.277>MLX, 00:15:45.332 --> 00:15:45.385 align:start position:0% Hugging Face Transformers, Apple MLX, 00:15:45.385 --> 00:15:46.598 align:start position:0% Hugging Face Transformers, Apple MLX, llama.cpp, <00:15:45.965>and <00:15:46.545>Unsloth, 00:15:46.598 --> 00:15:46.650 align:start position:0% llama.cpp, and Unsloth, 00:15:46.650 --> 00:15:48.285 align:start position:0% llama.cpp, and Unsloth, optimized <00:15:47.309>it <00:15:47.968>for 00:15:48.285 --> 00:15:48.338 align:start position:0% optimized it for 00:15:48.338 --> 00:15:50.078 align:start position:0% optimized it for AMD, <00:15:48.847>Apple, <00:15:49.356>Intel, <00:15:49.865>and 00:15:50.078 --> 00:15:50.131 align:start position:0% AMD, Apple, Intel, and 00:15:50.131 --> 00:15:51.607 align:start position:0% AMD, Apple, Intel, and Nvidia <00:15:50.341>chips, <00:15:50.551>and <00:15:50.761>you <00:15:50.971>can 00:15:51.607 --> 00:15:51.660 align:start position:0% Nvidia chips, and you can 00:15:51.660 --> 00:15:52.609 align:start position:0% Nvidia chips, and you can further 00:15:52.609 --> 00:15:52.662 align:start position:0% further 00:15:52.662 --> 00:15:54.719 align:start position:0% further customize <00:15:53.330>it <00:15:53.998>in <00:15:54.666>PyTorch. 00:15:54.719 --> 00:15:56.928 align:start position:0% customize it in PyTorch. 00:15:56.928 --> 00:15:59.284 align:start position:0% It <00:15:57.408>is <00:15:57.888>available <00:15:58.368>on <00:15:58.848>Hugging 00:15:59.284 --> 00:15:59.346 align:start position:0% It is available on Hugging 00:15:59.346 --> 00:16:01.578 align:start position:0% It is available on Hugging Face <00:15:59.718>under <00:16:00.090>the <00:16:00.462>name <00:16:00.834>Meta <00:16:01.206>Models 00:16:01.578 --> 00:16:01.640 align:start position:0% Face under the name Meta Models 00:16:01.640 --> 00:16:04.426 align:start position:0% Face under the name Meta Models Muse <00:16:02.135>Spark <00:16:02.630>Open <00:16:03.125>8 <00:16:03.620>16B42B. <00:16:04.115>Despite 00:16:04.426 --> 00:16:04.456 align:start position:0% Muse Spark Open 8 16B42B. Despite 00:16:04.456 --> 00:16:05.289 align:start position:0% Muse Spark Open 8 16B42B. Despite the <00:16:04.597>branding, <00:16:04.738>this <00:16:04.879>is <00:16:05.020>not <00:16:05.161>the 00:16:05.289 --> 00:16:05.319 align:start position:0% the branding, this is not the 00:16:05.319 --> 00:16:06.429 align:start position:0% the branding, this is not the strongest <00:16:05.479>open-weight <00:16:05.639>model <00:16:05.799>in <00:16:05.959>the <00:16:06.119>world 00:16:06.429 --> 00:16:07.569 align:start position:0% strongest open-weight model in the world 00:16:07.569 --> 00:16:07.600 align:start position:0% 00:16:07.600 --> 00:16:08.246 align:start position:0% . <00:16:08.141>In 00:16:08.246 --> 00:16:08.298 align:start position:0% . In 00:16:08.298 --> 00:16:09.930 align:start position:0% . In programming <00:16:09.720>and 00:16:09.930 --> 00:16:09.983 align:start position:0% programming and 00:16:09.983 --> 00:16:11.615 align:start position:0% programming and reasoning, <00:16:10.588>it <00:16:11.193>is 00:16:11.615 --> 00:16:11.667 align:start position:0% reasoning, it is 00:16:11.667 --> 00:16:13.457 align:start position:0% reasoning, it is inferior <00:16:12.246>to <00:16:12.825>Kimiko <00:16:13.404>3, 00:16:13.457 --> 00:16:13.510 align:start position:0% inferior to Kimiko 3, 00:16:13.510 --> 00:16:15.510 align:start position:0% inferior to Kimiko 3, Mimiko <00:16:13.765>V <00:16:14.020>2.6 <00:16:14.275>Pro, <00:16:14.530>Qwen <00:16:14.785>3.8, <00:16:15.040>Flash <00:16:15.295>Next 00:16:15.510 --> 00:16:15.562 align:start position:0% Mimiko V 2.6 Pro, Qwen 3.8, Flash Next 00:16:15.562 --> 00:16:17.215 align:start position:0% Mimiko V 2.6 Pro, Qwen 3.8, Flash Next and <00:16:15.929>GLM <00:16:16.296>5.3 <00:16:16.663>Flash. 00:16:17.215 --> 00:16:17.246 align:start position:0% and GLM 5.3 Flash. 00:16:17.246 --> 00:16:18.321 align:start position:0% and GLM 5.3 Flash. Agents' <00:16:17.476>performance <00:16:17.706>looks 00:16:18.321 --> 00:16:18.352 align:start position:0% Agents' performance looks 00:16:18.352 --> 00:16:19.502 align:start position:0% Agents' performance looks much <00:16:18.771>better. <00:16:19.190>According 00:16:19.502 --> 00:16:19.528 align:start position:0% much better. According 00:16:19.528 --> 00:16:20.337 align:start position:0% much better. According to <00:16:19.789>the <00:16:20.050>model <00:16:20.311>scorecard, 00:16:20.337 --> 00:16:20.363 align:start position:0% to the model scorecard, 00:16:20.363 --> 00:16:21.067 align:start position:0% to the model scorecard, it <00:16:20.597>actually 00:16:21.067 --> 00:16:21.093 align:start position:0% it actually 00:16:21.093 --> 00:16:21.901 align:start position:0% it actually outperforms <00:16:21.640>some 00:16:21.901 --> 00:16:21.927 align:start position:0% outperforms some 00:16:21.927 --> 00:16:22.657 align:start position:0% outperforms some models <00:16:22.135>with <00:16:22.343>higher 00:16:22.657 --> 00:16:22.683 align:start position:0% models with higher 00:16:22.683 --> 00:16:23.543 align:start position:0% models with higher overall 00:16:23.543 --> 00:16:23.569 align:start position:0% overall 00:16:23.569 --> 00:16:24.560 align:start position:0% overall intelligence <00:16:23.777>scores <00:16:23.985>on <00:16:24.193>certain 00:16:24.560 --> 00:16:24.586 align:start position:0% intelligence scores on certain 00:16:24.586 --> 00:16:25.678 align:start position:0% intelligence scores on certain agent <00:16:25.079>metrics. <00:16:25.572>Meta 00:16:25.678 --> 00:16:25.704 align:start position:0% agent metrics. Meta 00:16:25.704 --> 00:16:26.544 align:start position:0% agent metrics. Meta has <00:16:25.879>also <00:16:26.054>been <00:16:26.229>quite 00:16:26.544 --> 00:16:26.570 align:start position:0% has also been quite 00:16:26.570 --> 00:16:27.489 align:start position:0% has also been quite transparent <00:16:26.867>about <00:16:27.164>the <00:16:27.461>fact 00:16:27.489 --> 00:16:27.516 align:start position:0% transparent about the fact 00:16:27.516 --> 00:16:28.513 align:start position:0% transparent about the fact that <00:16:27.743>the <00:16:27.970>comparison <00:16:28.197>mixes 00:16:28.513 --> 00:16:28.540 align:start position:0% that the comparison mixes 00:16:28.540 --> 00:16:29.143 align:start position:0% that the comparison mixes tools, 00:16:29.143 --> 00:16:29.170 align:start position:0% tools, 00:16:29.170 --> 00:16:29.747 align:start position:0% tools, bindings, <00:16:29.642>and 00:16:29.747 --> 00:16:29.774 align:start position:0% bindings, and 00:16:29.774 --> 00:16:30.640 align:start position:0% bindings, and self-provided 00:16:30.640 --> 00:16:30.666 align:start position:0% self-provided 00:16:30.666 --> 00:16:31.218 align:start position:0% self-provided data <00:16:30.784>from <00:16:30.902>some 00:16:31.218 --> 00:16:31.244 align:start position:0% data from some 00:16:31.244 --> 00:16:32.344 align:start position:0% data from some vendors. <00:16:32.106>Muse 00:16:32.344 --> 00:16:32.404 align:start position:0% vendors. Muse 00:16:32.404 --> 00:16:34.731 align:start position:0% vendors. Muse Spark <00:16:32.926>Open <00:16:33.448>ran <00:16:33.970>on <00:16:34.492>Muse 00:16:34.731 --> 00:16:34.791 align:start position:0% Spark Open ran on Muse 00:16:34.791 --> 00:16:36.999 align:start position:0% Spark Open ran on Muse Code, <00:16:35.119>Kimiko <00:16:35.447>3 <00:16:35.775>on <00:16:36.103>Kimiko, <00:16:36.431>and <00:16:36.759>Qwen 00:16:36.999 --> 00:16:37.059 align:start position:0% Code, Kimiko 3 on Kimiko, and Qwen 00:16:37.059 --> 00:16:38.432 align:start position:0% Code, Kimiko 3 on Kimiko, and Qwen 3.8 <00:16:37.496>on <00:16:37.933>Claude <00:16:38.370>Code. 00:16:38.432 --> 00:16:38.607 align:start position:0% 3.8 on Claude Code. 00:16:38.607 --> 00:16:39.435 align:start position:0% 3.8 on Claude Code. The performance of 00:16:39.435 --> 00:16:39.465 align:start position:0% The performance of 00:16:39.465 --> 00:16:40.442 align:start position:0% The performance of agents <00:16:39.909>really 00:16:40.442 --> 00:16:40.471 align:start position:0% agents really 00:16:40.471 --> 00:16:41.152 align:start position:0% agents really depends <00:16:40.723>on <00:16:40.975>the 00:16:41.152 --> 00:16:41.182 align:start position:0% depends on the 00:16:41.182 --> 00:16:42.129 align:start position:0% depends on the combination <00:16:41.640>of <00:16:42.098>model, 00:16:42.129 --> 00:16:42.159 align:start position:0% combination of model, 00:16:42.159 --> 00:16:42.691 align:start position:0% combination of model, binding, 00:16:42.691 --> 00:16:42.721 align:start position:0% binding, 00:16:42.721 --> 00:16:43.461 align:start position:0% binding, tooling, 00:16:43.461 --> 00:16:43.491 align:start position:0% tooling, 00:16:43.491 --> 00:16:44.645 align:start position:0% tooling, context <00:16:44.053>management, 00:16:44.645 --> 00:16:44.675 align:start position:0% context management, 00:16:44.675 --> 00:16:45.415 align:start position:0% context management, and 00:16:45.415 --> 00:16:45.444 align:start position:0% and 00:16:45.444 --> 00:16:46.362 align:start position:0% and runtime. <00:16:46.334>So, 00:16:46.362 --> 00:16:46.391 align:start position:0% runtime. So, 00:16:46.391 --> 00:16:47.471 align:start position:0% runtime. So, open <00:16:46.817>models <00:16:47.243>should 00:16:47.471 --> 00:16:47.500 align:start position:0% open models should 00:16:47.500 --> 00:16:48.552 align:start position:0% open models should be <00:16:47.734>evaluated <00:16:47.968>not <00:16:48.202>just <00:16:48.436>by 00:16:48.552 --> 00:16:48.580 align:start position:0% be evaluated not just by 00:16:48.580 --> 00:16:49.633 align:start position:0% be evaluated not just by leaderboards, <00:16:49.576>but 00:16:49.633 --> 00:16:49.661 align:start position:0% leaderboards, but 00:16:49.661 --> 00:16:50.742 align:start position:0% leaderboards, but by <00:16:50.173>whether 00:16:50.742 --> 00:16:50.770 align:start position:0% by whether 00:16:50.770 --> 00:16:51.879 align:start position:0% by whether they <00:16:51.026>fit <00:16:51.282>into <00:16:51.538>real-world 00:16:51.879 --> 00:16:51.907 align:start position:0% they fit into real-world 00:16:51.907 --> 00:16:52.999 align:start position:0% they fit into real-world agent <00:16:52.861>workflows. 00:16:52.999 --> 00:16:53.033 align:start position:0% agent workflows. 00:16:53.033 --> 00:16:54.037 align:start position:0% agent workflows. That's <00:16:53.356>all <00:16:53.679>for <00:16:54.002>today. 00:16:54.037 --> 00:16:54.137 align:start position:0% That's all for today. 00:16:54.137 --> 00:16:55.237 align:start position:0% That's all for today. Write <00:16:54.483>your <00:16:54.829>thoughts <00:16:55.175>in 00:16:55.237 --> 00:16:55.269 align:start position:0% Write your thoughts in 00:16:55.269 --> 00:16:56.491 align:start position:0% Write your thoughts in the <00:16:55.635>comments. <00:16:56.001>Thanks <00:16:56.367>for 00:16:56.491 --> 00:16:56.522 align:start position:0% the comments. Thanks for 00:16:56.522 --> 00:16:57.691 align:start position:0% the comments. Thanks for watching, <00:16:56.799>see <00:16:57.076>you 00:16:57.691 --> 00:16:57.722 align:start position:0% watching, see you 00:16:57.722 --> 00:16:59.799 align:start position:0% watching, see you in <00:16:58.070>the <00:16:58.418>next <00:16:58.766>video.