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 Research & technical analysis Concern & critique Markets & investment
3 live reaction maps in this view Jump to the evidence ↓
Multiple readings Hugging face model evaluation security incident Concern & critique · 5 attributable reactions Reaction converging A global workspace in language models \ Anthropic Research & technical analysis · 6 attributable reactions Reaction converging Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident Concern & critique · 5 attributable reactions
Showing signals surfaced by Grace ×
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.

Established Evaluation & benchmarks Signal 6d 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 Models & releases Research 8d ago
⚡ 3 h early
global workspace theory language models paper

A global workspace in language models \ Anthropic

9 experts across 5 network communities independently surfaced this.

Why this matches Building & implementation reaction 1 attributable expert contribution · Ethan Mollick
“Some really interesting research from Anthropic that AI models have spontaneously developed a workspace that "appears to support the functions associated with conscious access" Demo of how this works: www.neuronpedia.org/qwen3.6-27b/... Research: www.anthro…” evidence ↗
9 experts 5 communities 1 sources clustered
Ethan Mollick Ryan Moulton Grace René Walter Tim Duffy Amy Hoy Michiel Bontenbal Stephan Geering +1
How the network is reacting 4 experts are independently emphasizing research & technical analysis.
Reaction converging

Research & technical analysis

4 experts

Evidence, methods and technical implications.

“Also we do look in their heads and they seem intelligent. www.anthropic.com/research/glo...”

Markets & investment

1 expert

Capital, companies and commercial impact.

“Anthropic has a new paper out, alleging a global workspace in LLMs. The term comes from Global Workspace Theory, a leading theory of consciousness. The method they use to investigate this is a refinement of logit lens, which they call J-lens. www.anthropic.…”

Building & implementation

1 expert

How teams are shipping and applying it.

“Some really interesting research from Anthropic that AI models have spontaneously developed a workspace that "appears to support the functions associated with conscious access" Demo of how this works: www.neuronpedia.org/qwen3.6-27b/... Research: www.anthro…”

2 experts discussed this · 25 posts
Ryan Moulton: Not non-physical, but non-functional. Despite the vast physical differences, we observe LLMs doing all the behaviors we typically believe require thinking.
Ryan Moulton: Not exactly turing-test, but in the same vein. If we see models solving the greatest intellectual challenges of humanity, using processes that look very much like internal monologue, either we have…
Ryan Moulton: And at the point that we exclude thinking as a prerequisite for solving all intellectual tasks, we've more or less eliminated its meaning entirely.
Open the full discussion →
Primary source anthropic.com ↗
New Agents & robotics Development 6h 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 3 experts are independently emphasizing concern & critique.
Reaction converging

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 ↗
More conversations

What experts are discussing without an anchoring article

15 experts · 41 posts · 13d ago
Mike Masnick Scott McGrath Grace Liz Fong-Jones (方禮真) jake Dan Hon Doll Alex Rubinsteyn
Matches Building & implementation
“A large amount of people are hating on Dame and Bailey for building the "bsky38" voting site that was going around the last few days, after they discovered that AI was used in building it, and espe…” evidence ↗
Multiple readings Concern & critique · 3 Research & technical analysis · 2 Policy & governance · 1
Grace: He's upfront with when he uses AI, so you can trust he's upfront with when he doesn't
mackuba.eu: Please just stop…
Open the thread →
6 experts · 18 posts · 3d ago
Anil Dash Grace jake Max Woolf Mary Branscombe mackuba.eu
Matches Building & implementation
“I think the whole point is that it's pretty hard to have a conversation about how to build social cohesion where a reply about "here's how LLMs learn" reads as a good-faith response.” evidence ↗
Multiple readings Building & implementation · 1 Questions & unknowns · 1
Anil Dash: Tressie mostly doesn't quote folks to dunk on them, fwiw
Anil Dash: I'm saying, she's quoted me plenty of times, it's a conversational style. I didn't read it as anything more than disagreeing with your point. At any rate, on the substance of what you're saying, I …
Open the thread →
5 experts · 35 posts · 15d ago
joao Grace Colin coty.bsky.social david-p-reichert.bsky.social
Matches Building & implementation
“The question, then, simply reduces to whether we will succeed at build something that smart. As Yudkowsky puts it, If Anyone Builds It, Everyone Dies. The question of how it will do this is kind of…” evidence ↗
Colin: Rationalists and adjacent folks, let me know if I've misrepresented anything! Stay tuned for another thread about why I think this is wrong.
Colin: Solving this is what's low key meant here by "solving alignment". Yes, factual hallucinations and whatnot are one problem. But the existential problem is that if we make this thing smart enough the…
Open the thread →
3 experts · 48 posts · 4d ago
Grace B! 🐝 Cavello (they/them) Singularity's Bounty e/cc
Matches Building & implementation
“What do we do about this? Substack lets you check for AI with pangram, but it doesn’t show a label upfront. I assume if it started doing that authors would optimize against it. Maybe build it into …” evidence ↗
Multiple readings Policy & governance · 1 Building & implementation · 1 Questions & unknowns · 1
Grace: AI writing, unfortunately
Grace: What do we do about this? Substack lets you check for AI with pangram, but it doesn’t show a label upfront. I assume if it started doing that authors would optimize against it. Maybe build it into …
Open the thread →