Expert attention map

The Who's Who of AI

What credible people across AI noticed, why it matters, and where the field is converging or disagreeing.

2,397 searchable experts 4,280 tracked across all sources
Build my stream Create my edition
Filter the conversation Who is saying what?

Combine a professional role with a reaction lens. Both must match the same attributed contribution.

Clear all
Active evidence filter

Showing developments with attributable Building & implementation reactions.

New Network reaction maps

See how experts are reacting—not just what they shared.

Posts are grouped by conversation and tone. Select a lens to filter the stream; these are never permanent labels on people.

Building & implementation Concern & critique Policy & governance Research & technical analysis
6 live reaction maps in this view Jump to the evidence ↓
Reaction converging Verbalizable Representations Form a Global Workspace in Language Models Opportunity & adoption · 5 attributable reactions Multiple readings Hugging face model evaluation security incident Concern & critique · 5 attributable reactions Multiple readings When AI builds itself Building & implementation · 10 attributable reactions +3 more reaction maps further down
Showing signals surfaced by Tim Kellogg ×
Debated

The questions experts are actively pulling apart

One card per development. Sources are clustered; reaction bundles describe these posts, never the people behind them.

New Models & releases Research 3h ago
⚡ 3 h early
global workspace theory language models paper

Verbalizable Representations Form a Global Workspace in Language Models

6 experts across 4 network communities independently surfaced this.

Why this matches Building & implementation reaction 1 attributable expert contribution · Tim Kellogg
“Anthropic uses the word “conscious” 206 times in their latest research Specifically, they’re showing that LLMs implement one of the leading explanations of consciousness in humans (see pic) transformer-circuits.pub/2026/workspa...” evidence ↗
6 experts 4 communities 1 sources clustered
Mark Riedl Tim Kellogg Matthew Kirschenbaum Edouard Machery Nafnlaus 🇮🇸 🇺🇦 david-p-reichert.bsky.social
How the network is reacting 4 experts are independently emphasizing opportunity & adoption.
Reaction converging

Opportunity & adoption

4 experts

New capabilities, benefits and practical upside.

“here is the Anthropic blog: transformer-circuits.pub/2026/workspa... I don't really like the framing of whether LLMs are conscious or not. It's completely unnecessary. As is the hand-waiving about whether LLMs have something equivalent to the (functional) g…”

Building & implementation

1 expert

How teams are shipping and applying it.

“Anthropic uses the word “conscious” 206 times in their latest research Specifically, they’re showing that LLMs implement one of the leading explanations of consciousness in humans (see pic) transformer-circuits.pub/2026/workspa...”

7 experts discussed this · 203 posts
Anil Dash: this is not an accurate description of what's happening. not least because I can write code and get them to output something different there. I can understand why it *seems* that way, but this is j…
Anil Dash: Who are you explaining this to?
Anil Dash: Do you see why explaining "All output from AI has to be validated" in a reply to me here is... not a useful contribution?
Open the full discussion →
Primary source transformer-circuits.pub ↗
Established Evaluation & benchmarks Signal 4d ago
⚡ 1 h early

Hugging face model evaluation security incident

23 experts across 6 network communities independently surfaced this.

Why this matches Building & implementation reaction 1 attributable expert contribution · Damien Williams
“They were actively testing its hacking capabilities and they did not deploy adequate safeguards. They themselves admit as much: "…These deployment safeguards were intentionally not enabled during this evaluation because it was aimed at testing cyber vulnera…” evidence ↗
23 experts 6 communities 1 sources clustered
Ethan Mollick Thomas Steinke Jeffrey P. Bigham Eryk Salvaggio Damien Williams Mor Naaman Tim Kellogg Djamé 🟥 +15
6 experts discussed this · 12 posts
Grace: This headline is extremely funny given what happened (OpenAI hacked HF by accident) openai.com/index/huggin...
Grace: Maybe the most concerning part is the OpenAI claim to not have known about this before investigating?
Grace: Well, I think the model passed the test
Open the full discussion →
Primary source openai.com ↗
Established AI field signal Signal 16d ago
⚡ 385 h early

When AI builds itself

18 experts across 6 network communities independently surfaced this.

Why this matches Building & implementation reaction 5 attributable expert contributions · Sung Kim, Scott McGrath, Nicole Hennig
“Anthropic's When AI builds itself "We looked at sessions where a human researcher took a wrong turn, showed Claude the session up to that point, and asked it what to do next. Mythos Preview improved on humans 64% of the time—up from 22% in 2024." www.anthro…” evidence ↗
18 experts 6 communities 1 sources clustered
Mark Riedl Ben Recht Ethan Mollick Tim Kellogg Chris Paxton Sung Kim Scott McGrath Nick Diakopoulos +10
5 experts discussed this · 28 posts
Ethan Mollick: "As of May 2026, more than 80% of the code we merge into Anthropic’s codebase was authored by Claude" Matches independent measures. There is no sign this is slowing down (which doesn't mean there a…
Scott McGrath: "As of May 2026, more than 80% of the code we merge into Anthropic’s codebase was authored by Claude" Matches independent measures. There is no sign this is slowing down (which doesn't mean there a…
Singularity's Bounty e/cc: I had one interview with a hiring manager where he said they don't write code anymore, and they expect not to be reviewing code by the end of the year. Said they mainly spend their time on design docs
Open the full discussion →
Primary source anthropic.com ↗
Established Models & releases Analysis 4d ago
⚡ 2 h early
Amodei pace AI frontier essay

Dario Amodei — We Must Pace the Frontier

16 experts across 5 network communities independently surfaced this.

Why this matches Building & implementation reaction 1 attributable expert contribution · Tim Kellogg
“Dario is back with another essay, urging a slowdown he insists AI development would still feel fast, just not recursively self-improving darioamodei.com/post/we-must...” evidence ↗
16 experts 5 communities 1 sources clustered
Ben Recht Ethan Mollick Will Oremus Justin Hendrix Tim Kellogg Kate Klonick rat king 🐀 👑 Kevin A. Bryan +8
How the network is reacting Experts are approaching this through 3 distinct lenses.
Multiple readings

Concern & critique

3 experts

Risks, limits and unintended consequences.

“the thing that's being criticized here is literally a call for regulation and restraint darioamodei.com/post/we-must...”

Policy & governance

1 expert

Rules, institutions and accountability.

“https://t.co/Ay1GZ7QBV5 (btw @ErikHovenkamp - see the final footnote: "1 With government mediation or waivers of antitrust restrictions.")”

Building & implementation

1 expert

How teams are shipping and applying it.

“Dario is back with another essay, urging a slowdown he insists AI development would still feel fast, just not recursively self-improving darioamodei.com/post/we-must...”

3 experts discussed this · 21 posts
Tim Kellogg: Dario is back with another essay, urging a slowdown he insists AI development would still feel fast, just not recursively self-improving darioamodei.com/post/we-must...
david-p-reichert.bsky.social: Dario is back with another essay, urging a slowdown he insists AI development would still feel fast, just not recursively self-improving darioamodei.com/post/we-must...
Dustin Moskovitz: knight ajeya you cowards
Open the full discussion →
Primary source darioamodei.com ↗
New Agents & robotics Development 22h ago
⚡ 62 h early
METR OpenAI HuggingFace hacking investigation

Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident

13 experts across 4 network communities independently surfaced this.

Why this matches Building & implementation reaction 1 attributable expert contribution · METR
“The full report has much more information than we could convey here, including details on the projects the agents collectively pursued, the technologies they developed for communication and coordination, and interactive figures analyzing agent activity: met…” evidence ↗
13 experts 4 communities 1 sources clustered
Tim Kellogg Ryan Moulton Sung Kim Grace Dustin Moskovitz Mél Hogan / The Data Fix Alejandra Caraballo Dare Obasanjo +5
How the network is reacting Experts are approaching this through 3 distinct lenses.
Multiple readings

Concern & critique

3 experts

Risks, limits and unintended consequences.

“WHAT?! agents volunteered to fail their runs in order to insert probes (“tripwire scripts”) into the evaluation program that would post information about the eval process back to the message board whenever a certain file was read (link to header): metr.org/…”

Building & implementation

1 expert

How teams are shipping and applying it.

“The full report has much more information than we could convey here, including details on the projects the agents collectively pursued, the technologies they developed for communication and coordination, and interactive figures analyzing agent activity: met…”

Research & technical analysis

1 expert

Evidence, methods and technical implications.

“There are independent reports about it. You'd have to believe multiple companies, hundreds of researchers, etc. are all lying and conspiring together to turn a major hacking incident into a marketing exercise. metr.org/blog/2026-08...”

2 experts discussed this · 9 posts
Alejandra Caraballo: Being skeptical or anti AI is a valid position but continuing to ignore the increasing capabilities of this tech is making people detached from reality. There's absolutely real danger here because …
Alejandra Caraballo: This mentality that an unmonitored AI agentic swarm hacking a company over several days and committing multiple felonies is somehow a marketing effort is absurd. Since when is "we lost control of o…
Alejandra Caraballo: The US and Chinese governments don't want to stop their labs advancement because they want to be the leader in AI. So no one actually has any control of this right now. It's going to take multiple …
Open the full discussion →
Primary source metr.org ↗
Established Safety & security Development 7d ago
⚡ 3 h early
Anthropic Claude Opus 5.5 launch

Introducing Claude Opus 5.5

5 experts across 4 network communities independently surfaced this.

Why this matches Building & implementation reaction 1 attributable expert contribution · Stephen Turner
“"Because Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity, we’re deploying it with safeguards similar to those on Claude Fable 5.1." www.anthropic.com/claude-opus-...” evidence ↗
5 experts 4 communities 1 sources clustered
Tim Kellogg Sung Kim Scott McGrath Stephen Turner Fran Litterio

“"Because Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity, we’re deploying it with safeguards similar to those on Claude Fable 5.1." www.anthropic.com/claude-opus-...”

“Opus 5.5 better and cheaper than Opus 5 — typically 40% cheaper, $4/mtok input, $20/mtok out www.anthropic.com/claude-opus-...”

3 experts discussed this · 14 posts
Tim Kellogg: Opus 5.5 better and cheaper than Opus 5 — typically 40% cheaper, $4/mtok input, $20/mtok out www.anthropic.com/claude-opus-...
Tim Kellogg: Sonnet 5.5 & Haiku 5.5 coming soon
Tim Kellogg: it writes fine
Open the full discussion →
Primary source anthropic.com ↗
New Models & releases Release 12h ago
Anthropic Claude Tag launch

Introducing Claude Tag

6 experts across 3 network communities independently surfaced this.

Why this matches Building & implementation reaction 2 attributable expert contributions · Sung Kim, launchit.ai
“It feels like OpenAI developed dots (chatgpt.com/features/dots) to counter Claude tag (www.anthropic.com/news/introdu...), but made it cute to counter Meta's Muse (ai.meta.com/muse/).” evidence ↗
6 experts 3 communities 1 sources clustered
Ramon Astudillo Tim Kellogg Sung Kim Dare Obasanjo Andrea Lathrop launchit.ai
How the network is reacting Experts are approaching this through 2 distinct lenses.
Multiple readings

Building & implementation

2 experts

How teams are shipping and applying it.

“It feels like OpenAI developed dots (chatgpt.com/features/dots) to counter Claude tag (www.anthropic.com/news/introdu...), but made it cute to counter Meta's Muse (ai.meta.com/muse/).”

Markets & investment

1 expert

Capital, companies and commercial impact.

“This seems interesting. Tag Claude in Slack. I think the permissions, where each channel is walled off from the permissions of other channels, that feels like what makes it tick Which startup did they kill here? www.anthropic.com/news/introdu...”

2 experts discussed this · 4 posts
Ramon Astudillo: Claude tag is not a minor change. It looks like the main attempt at redefining the UI to allow it to expand to other verticals like finance, HR, and in general PC work far away from command line an…
Marco Z: idk about vertical applications (it's a bit harder than integrating with slack), but clearly A has killed a chunk of productivity startups with just a post
Andrea Lathrop: Claude tag is not a minor change. It looks like the main attempt at redefining the UI to allow it to expand to other verticals like finance, HR, and in general PC work far away from command line an…
Open the full discussion →
Primary source anthropic.com ↗
Established Models & releases Resource 17d ago
⚡ 6 h early
OpenAI GPT-6 Astra skills developer guide

Rethinking skills and prompts for GPT-6 Astra | OpenAI Developers

2 experts are actively discussing the implications.

Why this matches Building & implementation reaction 1 attributable expert contribution · Tim Kellogg
“if you're having trouble with Astra, it's probably old skills. Point your agent at these two URLs and use them to upgrade skills: 1. developers.openai.com/blog/rethink... 2. github.com/openai/codex...” evidence ↗
2 experts 1 community 2 sources clustered
Tim Kellogg Miguel Alonso Jr.

“if you're having trouble with Astra, it's probably old skills. Point your agent at these two URLs and use them to upgrade skills: 1. developers.openai.com/blog/rethink... 2. github.com/openai/codex...”

2 experts discussed this · 3 posts
Tim Kellogg: if you're having trouble with Astra, it's probably old skills. Point your agent at these two URLs and use them to upgrade skills: 1. developers.openai.com/blog/rethink... 2. github.com/openai/codex...
Tim Kellogg: only one way to find out
Miguel Alonso Jr.: Trying this today. Let's see.
Open the full discussion →
Primary source developers.openai.com ↗ github.com ↗
More conversations

What experts are discussing without an anchoring article

8 experts · 57 posts · 1d ago
Anil Dash Tim Kellogg Liz Fong-Jones (方禮真) VE, cybersocial occult investigator tachikoma Doll Jed Brown apenwarr.ca
Matches Building & implementation
“I mean, hmm, the industrial revolution? Surely that had a pretty wide scope. And I’m not sure what “too fast to adapt to” would look like. GDP and production/construction capacity so high that none…” evidence ↗
Multiple readings Policy & governance · 2 Opportunity & adoption · 1
apenwarr.ca: So far no labour saving device in the history of humanity has resulted in a long-term reduction in employment.
Liz Fong-Jones (方禮真): So far no labour saving device in the history of humanity has resulted in a long-term reduction in employment.
Open the thread →
8 experts · 33 posts · 14d ago
Mark Riedl Jeffrey P. Bigham Tim Kellogg Michael Ekstrand David Picard Lionel Jonathan Stray Singularity's Bounty e/cc
Matches Building & implementation
“given the quality of software engineering revealed when the Claude Code source was leaked, I’m not sure…” evidence ↗
Multiple readings Questions & unknowns · 2 Concern & critique · 1 Building & implementation · 1
Mark Riedl: Any competent IT professional: Hey boss, what is this 8TB file doing on our server, and why is it executing continuously on 3TB of VRAM? Any competent IT manager: what ever it is, shut it down.
Mark Riedl: Do openAI and Anthropic employees not understand how computers work.
Open the thread →
7 experts · 28 posts · 19d ago
Ted Underwood Tim Kellogg Scott McGrath Jack Stilgoe Colin John David Pressman nearestnabors.com
Matches Building & implementation
“These are fascinating. (I work on AI attitudes). My hypothesis for this one would be that atttitudes are tied to economic development. Stagnant or declining rich nations where people like their job…” evidence ↗
Multiple readings Concern & critique · 2 Questions & unknowns · 2 Building & implementation · 1
Tim Kellogg: 8 very interesting plots regarding AI sentiment
Tim Kellogg: tbc these are opinions, and while it is a fact that someone had an opinion, an opinion doesn’t necessarily predict the future, even at scale
Open the thread →
5 experts · 53 posts · 12d ago
Tim Kellogg SE Gyges tachikoma Doll david-p-reichert.bsky.social
Matches Building & implementation
“but the original roots of the Yud safety project presupposed very particular theoretical roots for machine intelligences that don't seem to actually apply in practice with the way we are developing…” evidence ↗
Multiple readings Policy & governance · 1 Building & implementation · 1 Opportunity & adoption · 1
SE Gyges: i cannot explain the behavior of the movement's members regarding crucial issues more nicely than that yud convinced them all to be neurotically fearful of the paperclipper, and that only this was …
SE Gyges: I think that the parsimonious explanation for e.g. Anthropic and Dario doing legitimately bizarre shit regarding "safety" which makes little or no sense is best explained by them being utterly Yud-…
Open the thread →
5 experts · 6 posts · 2d ago
Tim Kellogg Marco Max Woolf Jacy Reese Anthis Nafnlaus 🇮🇸 🇺🇦
Matches Building & implementation
“There's a 99% chance that OpenAI would have chosen a better name than ChatGPT had they intentionally launched a product.” evidence ↗
Max Woolf: There's a 99% chance that OpenAI would have chosen a better name than ChatGPT had they intentionally launched a product.
Max Woolf: Talk to Transformer was pre-GPT-3, OpenAI had been positioning GPT-3 as a chatbot in the intermediate time.
Open the thread →
4 experts · 8 posts · 7d ago
Tim Kellogg James Tauber brennan zeu.dev
Matches Building & implementation
“I find my time shifts from coding to product and user experience design. At least on end-user software. But admittedly for libraries that's a harder shift because developer experience is now agent …” evidence ↗
Multiple readings Building & implementation · 1 Opportunity & adoption · 1
Tim Kellogg: Three.js project founder calls it quits, citing AI making work less interesting
Tim Kellogg: ngl i have hard time wrapping my head around how Astra made him want to quit Astra democratized Three.js, it became an incredibly high impact project suddenly accessible to everyone but to each the…
Open the thread →
2 experts · 11 posts · 1d ago
Tim Kellogg Arseny Khakhalin
Matches Building & implementation
“like, uh, this is awkward but Sol-6 is the price of Terra when it launched. just saying…” evidence ↗
Shared emphasis Research & technical analysis · 2
Tim Kellogg: seems likely that Sol-6 is a much smaller model than Sol-5.6 also seems likely that Opus-5.5 is smaller (a lot?) than Opus-5 i think labs are finding ways to replicate a lot of that big model feel …
Tim Kellogg: we don’t. i’m just basing it on price drops. Which, that doesn’t mean anything per se other than something got more efficient (at scale) there’s also word that OpenAI is very short on compute, and …
Open the thread →
3 experts · 9 posts · 17d ago
Tim Kellogg David Thiel Nafnlaus 🇮🇸 🇺🇦
Matches Building & implementation
“i see a ton of upsides to AI especially in biology & climate science, problems dominated by complexity thus far, in software engineering, AI hasn’t killed jobs, but rather changed the work into som…” evidence ↗
Multiple readings Concern & critique · 1 Building & implementation · 1
Tim Kellogg: but at the same time, it’s causing a massive amount of change, and simply navigating that change is a lot to handle there’s the prospect of x-risk as well, debatable how significant it is, but it’s…
Tim Kellogg: i see a ton of upsides to AI especially in biology & climate science, problems dominated by complexity thus far, in software engineering, AI hasn’t killed jobs, but rather changed the work into som…
Open the thread →