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
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.

Research & technical analysis Concern & critique Building & implementation Markets & investment
6 live reaction maps in this view Jump to the evidence ↓
Multiple readings OpenAI’s latest math breakthroughs commit research misconduct, experts say Research & technical analysis · 2 attributable reactions Reaction converging Models Don't Go Rogue Concern & critique · 3 attributable reactions Reaction converging IN THE WEIGHTS Research & technical analysis · 4 attributable reactions +3 more reaction maps further down
Showing signals surfaced by René Walter ×
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 Development 21h ago
OpenAI math research misconduct allegations

OpenAI’s latest math breakthroughs commit research misconduct, experts say

8 experts across 5 network communities independently surfaced this.

8 experts 5 communities 1 sources clustered
Emily M. Bender Timnit Gebru joao Olivia Guest BostonJoan 🏳️‍🌈 🦫 Eileen Clancy 🧿 Tommy Löfstedt René Walter
How the network is reacting Experts are approaching this through 2 distinct lenses.
Multiple readings

Research & technical analysis

1 expert

Evidence, methods and technical implications.

“OpenAI's latest math breakthroughs commit research misconduct, experts say archived: archive.ph/JCl4g”

Questions & unknowns

1 expert

What remains unresolved or contested.

“OpenAI claimed their products have solved several complex math problems, so what’s the issue? OpenAI just took credit for work done by humans without proper citations. When people in my field do that, they get tenure. Hahaha Red more: www.scientificamerican…”

6 experts discussed this · 48 posts
Timnit Gebru: These men who say factually and technically wrong things with such confidence are so fascinating to me. HE doesn't know what a language model is and thinks that OUR formulation is wrong. Because HE…
Timnit Gebru: Emily Bender is a more patient person than me and has written a Frequently Unasked Questions about this. I have seen a whole bunch of people saying this like its some gotcha talking point they've a…
Timnit Gebru: We even DEFINED what language models are in our paper and gave their history.
Open the full discussion →
Primary source scientificamerican.com ↗
Established Safety & security Analysis 3d ago
models don't go rogue essay

Models Don't Go Rogue

8 experts across 3 network communities independently surfaced this.

8 experts 3 communities 1 sources clustered
Timnit Gebru Eryk Salvaggio Luke Stark Shannon Mattern Mél Hogan / The Data Fix Benjamin Riley Elke Schwarz René Walter
How the network is reacting 3 experts are independently emphasizing concern & critique.
Reaction converging

Concern & critique

3 experts

Risks, limits and unintended consequences.

“Well we just disagree here, that's ok. Because I don't think it helps anyone to credit AI agents with "building a civilization" (or three) and hinting that the little guys getting dangerously smart. I'm with @eryk.bsky.social: mail.cyberneticforests.com/mod…”

5 experts discussed this · 20 posts
Timnit Gebru: Friends, when I sent this message to my team laughing, I didn't know that this fairytale blog had been shared far & wide & that people actually believed it. I didn't know that until someone asked m…
Timnit Gebru: At a certain point I have to ask myself if I got all this pedigree and spent decades doing research and building things to then spend the rest of my life attempting to come up with good arguments a…
Timnit Gebru: Agents didn't "confer" that is anthropomorphization of computer programs trained to do certain tasks.
Open the full discussion →
Primary source mail.cyberneticforests.com ↗
Established AI research Resource 19d ago
⚡ 31 h early
In the Weights newsletter issue

IN THE WEIGHTS

12 experts across 5 network communities independently surfaced this.

12 experts 5 communities 1 sources clustered
Lynn Cherny Kashmir Hill Sasha Costanza-Chock Reuben Binns Dare Obasanjo Ben Williamson Meredith Martin Aram Sinnreich +4
How the network is reacting 3 experts are independently emphasizing research & technical analysis.
Reaction converging

Research & technical analysis

3 experts

Evidence, methods and technical implications.

“are you in the weights? Hilarious and fascinating including the mess ups intheweights.com”

Concern & critique

1 expert

Risks, limits and unintended consequences.

“LLMs never fail to simultaneously impress and disappoint. Two searches for my name at intheweights.com showing completely different results with one accurate and one containing hallucinations.”

4 experts discussed this · 6 posts
austin: fun? I am in there. intheweights.com
VE, cybersocial occult investigator: Well, my strategy of having an incredibly common wallet name is still rendering me invisible to information systems.
austin: weirdly enough my github handle actually had a better score than my wallet name
Open the full discussion →
Primary source intheweights.com ↗
Established Agents & robotics Development 4d 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

10 experts across 4 network communities independently surfaced this.

10 experts 4 communities 1 sources clustered
Tim Kellogg Ryan Moulton Sung Kim Grace Mél Hogan / The Data Fix Alejandra Caraballo Liz Fong-Jones (方禮真) Fran Litterio +2
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: 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: 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: There needs to be a global moratorium on frontier research for at least a few months if not a year while safeguards and safety research catches up. The problem is that no one has that ability. The …
Open the full discussion →
Primary source metr.org ↗
Developing AI field signal Signal 1d ago

AI may not make us stupid, but relying on it could weaken the skills built through practice

3 experts are actively discussing the implications.

1 expert 1 community 1 sources clustered
René Walter

“Latest i'm aware of found that cognitive offloading depends on usage style, basically if you use it as guidance for learning or as a text generator for results, which seems very obvious to me. No study on the general outcome afaik.”

3 experts discussed this · 15 posts
Shannon Vallor: bangs head slowly against wall
Justin Hendrix: Looking forward to your thoughts on the study
Shannon Vallor: It is not as informative as a longitudinal study like the recent one on Chinese undergrads that tracked performance of the same students over time, since this study can't say whether the students w…
Open the full discussion →
Primary source phys.org ↗
Established Models & releases Research 7d ago
⚡ 3 h early
global workspace theory language models paper

A global workspace in language models \ Anthropic

8 experts across 4 network communities independently surfaced this.

8 experts 4 communities 1 sources clustered
Ethan Mollick Grace René Walter Tim Duffy Amy Hoy Michiel Bontenbal Stephan Geering david-p-reichert.bsky.social
How the network is reacting 3 experts are independently emphasizing research & technical analysis.
Reaction converging

Research & technical analysis

3 experts

Evidence, methods and technical implications.

“LLMs similarly have nonphenomenal access to information, and can reflect and use memory. This is indeed similar in principle, while the cognitive architecture is vastly less complex (yet), and likely won't catch up anytime soon. Add to this Anthropics j-spa…”

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 · 4 posts
Tim Duffy: 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 refi…
Tim Duffy: More detail on the method from the paper:
Grace: 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 refi…
Open the full discussion →
Primary source anthropic.com ↗
Established AI field signal Analysis 9d ago
Cosma Shalizi internet traditions essay

Cosma Shalizi Is Aware of All Internet Traditions

4 experts are actively discussing the implications.

1 expert 1 community 1 sources clustered
René Walter

“2: This is the narratology version of Cosma Shalizis take: "By design, language models mechanistically reproduce the recurring regularities in their training data. That training data consists of all the text files on the internet and what is easily availabl…”

4 experts discussed this · 10 posts
René Walter: 1
Colin: Genre: more and more people are talking about it
René Walter: 2: This is the narratology version of Cosma Shalizis take: "By design, language models mechanistically reproduce the recurring regularities in their training data. That training data consists of al…
Open the full discussion →
Primary source argmin.net ↗
Established Models & releases Research 14d ago
⚡ 953 h early
State media influence LLM training paper

State media control influences large language models | Nature

8 experts across 3 network communities independently surfaced this.

8 experts 3 communities 1 sources clustered
Justin Hendrix Luke Stark Gina Helfrich Dr. Chinasa T. Okolo Debora Nozza MilaNLP Lab René Walter B! 🐝 Cavello (they/them)
How the network is reacting Experts are approaching this through 2 distinct lenses.
Multiple readings

Policy & governance

2 experts

Rules, institutions and accountability.

“"...LLMs exhibit a stronger pro-government valence in the languages of countries with lower media freedom than in those with higher media freedom." www.nature.com/articles/s41...”

Research & technical analysis

2 experts

Evidence, methods and technical implications.

“Researchers from Oregon, Purdue, UC San Diego, NYU and Princeton ran six experiments. The overarching finding: state control of media in many countries is already baked into the training data of widely used commercial models, and is shaping the answers thos…”

2 experts discussed this · 2 posts
VE, cybersocial occult investigator: Starting a conspiracy theory that the papal encyclical was actually trying to get ahead of and cement Catholicism as the official machine religion
René Walter: OMG www.nature.com/articles/s41...
Open the full discussion →
Primary source nature.com ↗
More conversations

What experts are discussing without an anchoring article

14 experts · 130 posts · 14h ago
Marc Lanctot Olivia Guest Sarah T Roberts, PhD Pwnallthethings Mike Caulfield Colin Adam Wiemerslage rev. howard arson
Multiple readings Questions & unknowns · 4 Opportunity & adoption · 2 Concern & critique · 1
Sarah T Roberts, PhD: Try teaching college students. You won’t be quite so dismissive then.
Pwnallthethings: I think the "LLMs can't reason" stuff is mostly sophism these days. They plainly do not "reason" in the way humans do, and are certainly not always correct or good at it, but they are also doing th…
Open the thread →
3 experts · 23 posts · 1d ago
Benjamin Riley Colin René Walter
Multiple readings Opportunity & adoption · 1 Questions & unknowns · 1
Colin: Huh? I don’t think either of these is true. I don’t think LLMs are indistinguishable from sentient things, and I also never thought sentience was a prerequisite for code completion or theorem provi…
Colin: I don’t think sentient beings have a “regenerate response” button. I don’t think their states are fully determined by a handful of tokens. They live in the world and navigate space and time like th…
Open the thread →
3 experts · 17 posts · 19h ago
Colin René Walter Singularity's Bounty e/cc
Multiple readings Research & technical analysis · 1 Context & explanation · 1
René Walter: That's the crux of all these debates, innit. How can we talk about quasi-cognite things when our words for cognite things are themselves only quasi-symbols with fuzzy meaning. A stone is a stone, a…
Colin: One take on thinking which I think is perfectly fine is just: thinking is the thing your brain does when you’re alive, or perhaps only when you’re awake. Clearly LLMs don’t do this: they don’t have…
Open the thread →
4 experts · 10 posts · 8d ago
Ted Underwood Ethan Mollick Sung Kim René Walter
Multiple readings Markets & investment · 1 Building & implementation · 1
Ethan Mollick: Kind of surprised that we are not seeing more radical political ideas built around AI capabilities, the way that early mass industrialization was a prime motivator in the conceptions of capitalism,…
René Walter: Bit of a myopic lense here, all those ideologies already have automatization baked in, and stuff like fully automated luxury communism just seems like another iteration.
Open the thread →
3 experts · 6 posts · 1d ago
Mike Caulfield Mary Branscombe René Walter
Mike Caulfield: The one comment I'll make is when everybody gets to choose the worst proponents of ideas to argue with everybody gets to be right and vindicated. And one problem is the worst proponents of the idea…
Mike Caulfield: I know a lot of people who are quite bullish on AI but I have never in my life met a single person (even in an online way) that sees a single thing that Altman or Huang says as anything other than …
Open the thread →
2 experts · 13 posts · 1d ago
Pwnallthethings René Walter
René Walter: What is the difference between "reasoning" and "reasoning-mimicry"? I'm not sure if that difference exists in a thing that sends language objects back and forth.
René Walter: roon on the other site recently wrote that it's not clear how useful the non-anthropomorphical way to describe LLMs is if the behavior LLMs show is itself anthropomorphic. How useful, or harmful, i…
Open the thread →
2 experts · 21 posts · 11d ago
Matthew Kirschenbaum René Walter
Shared emphasis Questions & unknowns · 2
René Walter: Are interpolatable archives plugged into each other communicating?
René Walter: True communication happens intersubjectively, and AI agents are not subjects with intent. There must be an "I" transmitting an internal reality to a "You". So, what are thousands of AI agents doing…
Open the thread →
2 experts · 15 posts · 16d ago
Andy Masley René Walter
Andy Masley: Well yes I agree this leaves open the chance that this could be about AI, but I think it's also evidence against the idea that this is a general backlash against the four companies actually buildin…
Andy Masley: Yeah but have you looked at the specific reasons why in this link? Only 14% say their opposition is driven by negative views of AI
Open the thread →
3 experts · 4 posts · 8d ago
Hagen Blix René Walter Amy Hoy
Hagen Blix: I don't think we're gonna win until we figure out how to be better story tellers than the other side. Cause.. you know, that sounds like a cool world to be living in where Dwarkesh presumably gets …
Amy Hoy: no, we don’t. we have to educate people, in a way that’s easy and fun to read i’ve explained how llms actually work to a lot of people, and had it change their attitude. they’re often both more int…
Open the thread →
2 experts · 4 posts · 5d ago
Benjamin Riley René Walter
Benjamin Riley: While getting my hair cut, my stylist — also in grad school — tells me she had a guest professor who told her class they need “AI literacy,” which apparently meant having Chat write biographies for…
René Walter: Obviously a generative biography is useful to show shortcomings of AI precisely because you know who you are. AI literacy is not a prompting class, it's about showing the process, and that's a good…
Open the thread →