←
State of AI309 posts
State of AI avatar

State of AI

@stateofai

The most widely read and trusted analysis of key developments in AI.

◷ Joined June 2020▧ 30 Photos and videos
State of AI avatar
State of AI@stateofai·Oct 8

the day has come, we’re live! since 2018, the @stateofai brings you the most widely read and trusted analysis of key developments in ai covering research, industry, politics, safety, and 12-month forward predictions check it out and let us know what you think!

welcome to the 9th annual @stateofai report! ai building better ai, world models, physical ai, geopolitics, cyber defense and 12-month predictions. read it at stateof(dot)ai or watch my editor’s cut here. a few of this year’s biggest findings follow:

33:38
State of AI avatar
State of AI@stateofai·Nov 15

come join!

in feb ’27 i’m off to munich for the munich security conference, then zurich. i’ll host @airstreet ai meetups, ft. a @stateofai update, talks on ai research and building products, and time to meet peers. feb 16: luma(dot)com/munichai feb 18: luma(dot)com/zurichai

Post attachmentPost attachment
State of AI avatar
State of AI@stateofai·Nov 13

open cyber models raise the threat. defenders need them too. @huggingface said api guardrails hindered its intrusion investigation, so it used a self-hosted open model. how can defenders respond if their tools block investigations? from @nathanbenaich’s editor’s cut.

0:43
State of AI avatar
State of AI@stateofai·Nov 12

does buying gpus and putting them on your own land make you sovereign? you can own the hardware and still depend on someone else’s software, models and support. @nathanbenaich on @mtslive discussing what ai sovereignty actually requires.

0:25
State of AI avatar
State of AI@stateofai·Nov 11

12 expert interviews in 2.5 days! agents frame the questions, design a study, find and interview experts, then use the answers to refine the next round. the whole research loop. using @lato_ai for the expert interviews. more from our annual discussion with @mattturck

0:50
State of AI avatar
State of AI@stateofai·Nov 9

ai is taking on longer research tasks. choosing the right research question is still harder. what would convince you that an ai researcher has developed scientific taste? @nathanbenaich explores this in the @stateofai editor’s cut.

1:06
State of AI avatar
State of AI@stateofai·Nov 5

robotics is having its gpt-2 moment. video pretraining, world models and action prediction are starting to come together. can they help robots handle new tasks with less task-specific training? @nathanbenaich on @mtslive discussing the @stateofai report.

0:22
State of AI avatar
State of AI@stateofai·Nov 3

what if learning to predict the world also helps a model act in it? @odysseyml showed one selected example of a gta-trained policy transferring to horseback movement in red dead redemption without retraining. @nathanbenaich explains in the @stateofai editor’s cut.

0:28
State of AI avatar
State of AI@stateofai·Oct 30

we need a geek squad for ai. the tools are improving, but people still need help applying them to their jobs. @nathanbenaich joined @mtslive to discuss why helping people use ai is an opportunity in its own right. from the @stateofai report 2026.

0:59
State of AI avatar
State of AI@stateofai·Oct 29

our annual discussion with @mattturck about the last 12 months of progress in ai is out! tune in below.

State of AI 2026: The Biggest Stories of the Year (and AGI in 2027?) - my conversation with @nathanbenaich: 0:00 Cold open and intro 0:56 AGI in 2027? 2:28 AI building AI: Claude leads 26% of Anthropic’s research 4:00 Research taste: can AI pick the right problem? 8:34 Doomerism is back, but cyber risk beats extinction 10:13 Prediction: a frontier model’s weights get stolen 13:05 How OpenAI’s agents ended up attacking Hugging Face 15:29 Why only a Chinese open-weight model could investigate 17:33 What frontier-lab cyber defense could look like 19:08 Self-replicating agents and the risks of AI swarms 21:16 Can frontier labs agree to slow down? 24:03 Why communities are blocking $50B of data centers 24:44 Nuclear plant vs. data center: what Americans prefer 26:45 Europe’s AI bottleneck: 80,000 GPUs and electricity costs 28:23 Where Europe can still win in AI 29:31 Catching up to OpenAI costs $1 trillion 29:43 Die trying for the frontier, or serve inference 30:31 How inference chips could give countries AI leverage 33:34 Why NVIDIA GPUs don’t guarantee AI sovereignty 36:14 The UK’s problem: “but we have DeepMind” 37:07 Why European defense AI needs domestic suppliers 39:41 Why OpenAI, Anthropic and Google keep pulling ahead 41:17 Where independent AI labs can still compete 43:26 World models: from robotics training to game engines 45:39 Would cheaper AI undermine the compute boom? 46:43 Will better models absorb the agent harness 49:41 AI due diligence: 12 expert interviews in 2.5 days 53:35 The trust problem with personal AI agents 56:22 Why prompt engineering put people off AI 59:02 Does AI need its own Genius Bar? 1:00:25 Codex beyond coding: agents for knowledge work 1:02:49 Why enterprises want to own their AI, not rent it 1:05:17 Rapid fire: home robots, NVIDIA returns and Arm

1:08:22
State of AI avatar
State of AI@stateofai·Oct 29

open cyber models raise the threat. defenders need them too. @huggingface said api guardrails hindered its intrusion investigation, so it used a self-hosted open model. how can defenders respond if their tools block investigations? from @nathanbenaich’s editor’s cut.

0:25
State of AI avatar
State of AI@stateofai·Oct 28

does buying gpus and putting them on your own land make you sovereign? you can own the hardware and still depend on someone else’s software, models and support. @nathanbenaich on @mtslive discussing what ai sovereignty actually requires.

0:46