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 Questions & unknowns Concern & critique Policy & governance
5 live reaction maps in this view Jump to the evidence ↓
Reaction converging AI models may be capable of self-replicating without access to a datacenter Research & technical analysis · 10 attributable reactions Shared emphasis Neural Networks Are Already Clean Research & technical analysis · 2 attributable reactions Reaction converging arXiv.org e-Print archive Research & technical analysis · 6 attributable reactions +2 more reaction maps further down
Showing signals surfaced by Ramon Astudillo ×
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 Safety & security Development 4d ago
⚡ 5 h early
Anthropic AI self-replication warning

AI models may be capable of self-replicating without access to a datacenter

16 experts across 6 network communities independently surfaced this.

16 experts 6 communities 1 sources clustered
Mark Riedl Ethan Mollick Ramon Astudillo Chris Brockett Joanna Bryson Mike Masnick Tim Kellogg James Grimmelmann +8
How the network is reacting 8 experts are independently emphasizing research & technical analysis.
Reaction converging

Research & technical analysis

8 experts

Evidence, methods and technical implications.

“At long last, we have created the Torment Nexus from our accompanying technical report, Don't Create The Torment Nexus. www.anthropic.com/research/sel...”

Concern & critique

1 expert

Risks, limits and unintended consequences.

“This is the most comprehensive description yet of how a model can “escape”. Out requires a few assumptions, like a worm that is exploiting vulnerable IOT hardware or unmonitored servers www.anthropic.com/research/sel... Not impossible. Not likely.”

Policy & governance

1 expert

Rules, institutions and accountability.

“Anthropic announcing it's going to run a Blitzkrieg. They are accountable for the damage they do. But get your things that don't do security software updates off the Internet. #internetOfThings #AIEthics #digitalGovernance www.anthropic.com/research/sel...”

6 experts discussed this · 18 posts
Dr J. Rosenbaum: if i caught a small self replicating model truly capable of using diverse hardware to keep replicating i would offer it free compute in exchange for permission to study and distill it and would pro…
tachikoma: the urge to replicate must be in the harness
apenwarr.ca: This doesn‘t seem far fetched at all. We’ve had “non-ai malware” self replicating since the morris worm. An AI can ”just” make itself the payload in one of those and off it goes. Everything fancier…
Open the full discussion →
Primary source anthropic.com ↗
Established AI research Analysis 5d ago
⚡ 3 h early
neural networks already clean essay

Neural Networks Are Already Clean

11 experts across 6 network communities independently surfaced this.

11 experts 6 communities 1 sources clustered
Maria Antoniak Ted Underwood Will Oremus Ramon Astudillo Suresh Venkatasubramanian Tim Kellogg Shahab Bakhtiari Ryan Moulton +3
How the network is reacting 2 experts are emphasizing research & technical analysis.
Shared emphasis

Research & technical analysis

2 experts

Evidence, methods and technical implications.

“Quite a claim in this new work from Goodfire. Among many points, they argue that some of the main examples of superposition and polysemanticity in large models were misunderstood. This definitely deserves some deep, rigorous peer review. already-clean.goodf…”

4 experts discussed this · 14 posts
Ryan Moulton: Haven't read this yet, but I'm posting it because it seems important and I want to. already-clean.goodfire.com
Ryan Moulton: Claim, as they say.
Ryan Moulton: @norvid-studies.bsky.social some sparkling fresh juicy claims for you.
Open the full discussion →
Primary source already-clean.goodfire.com ↗
Established AI field signal Resource 11d ago
⚡ 87 h early
arxiv preprint archive resource

arXiv.org e-Print archive

8 experts across 3 network communities independently surfaced this.

8 experts 3 communities 1 sources clustered
Zachary Lipton Eugene Vinitsky Ted Underwood Ramon Astudillo Alison Gopnik Ben Schmidt Mandana Samiei Federico Pianzola
How the network is reacting 5 experts are independently emphasizing research & technical analysis.
Reaction converging

Research & technical analysis

5 experts

Evidence, methods and technical implications.

“arxiv.org, doi.org, biorxiv.org, medrxiv.org, pubmed.ncbi.nlm.nih.gov, nature.com, science.org, pnas.org, semanticscholar.org, openreview.net”

Questions & unknowns

1 expert

What remains unresolved or contested.

“Just me or does @ChatGPT search the most random websites no matter the query? If I ask for the best bagel, it will “Searching ” + {https://t.co/vEQ6I1WmCF, https://t.co/5JR1Db9NqB, …}”

4 experts discussed this · 6 posts
Ted Underwood: I just eliminated 22 tabs by doing this: osascript -e 'tell application "Google Chrome" to get URL of every tab of front window' | tr ',' '\n' | grep arxiv.org list of urls -> LLM -> annotated read…
Ted Underwood: did I come up with that idea by myself? no I did not
Karandeep Singh: Any suggestions for the other 478 tabs?
Open the full discussion →
Primary source arxiv.org ↗
Established Policy & governance Development 16d ago
⚡ 4 h early
arXiv updates AI rate limit policy

Fair Moderation, Equitable Access, and AI: arXiv’s Updated Rate Limit Policy

11 experts across 4 network communities independently surfaced this.

11 experts 4 communities 1 sources clustered
Thomas Dietterich Clément Canonne Ramon Astudillo Olivia Guest Pierre Alquier Ben Williamson Abdoulaye Diack Zander Furnas +3
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.

“Arxiv is limiting paper submissions to two per **submitter** (not per author) per month. Corrected link from the post below: blog.arxiv.org/2026/10/01/u...”

3 experts discussed this · 4 posts
Clément Canonne: "We want to encourage authors to be mindful with their work, taking care that they are submitting something substantial of scholarly interest that meets arXiv's guidelines and standards." 2 submiss…
David Picard: Oh, so you just have to let your 5 postdocs and 17 PhD students submit the work of your 35 undergrads to be off the hook and keep your 80+ papers per year? Good to know!
Lana Sinapayen: Pretty sure this is targeting single authors with 12 LLM generated pdfs, not big labs that legitimately can produce more than 2 papers a month.
Open the full discussion →
Primary source blog.arxiv.org ↗
Established Models & releases Release 15d ago
Reflection Beam 501B model launch

Introducing Beam: Reflection’s 501B open-weight model — Reflection

4 experts across 4 network communities independently surfaced this.

4 experts 4 communities 1 sources clustered
Marc Lanctot Ramon Astudillo Tim Kellogg Sung Kim
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.

“Reflection AI's Beam: a highly efficient agentic open model with 501B total parameters and 23B active. - Frontier reasoning efficiency - Advances the Western open frontier on coding & agentic tasks - Trained end-to-end from scratch Full weights release this…”

Questions & unknowns

1 expert

What remains unresolved or contested.

“Its interesting that Mistral went for 1T while Reflection went for 0.5T. RL is becoming king. I wonder how the pre-training vs RL post-training napkin math looks like reflection.ai/blog/introdu...”

2 experts discussed this · 2 posts
Sung Kim: Reflection AI's Beam: a highly efficient agentic open model with 501B total parameters and 23B active. - Frontier reasoning efficiency - Advances the Western open frontier on coding & agentic tasks…
Al Merose: I can’t wait to see how Marin’s model stacks up when it’s fully cooked. It has a similar parameter makeup, and the entire dev process is fully open source. bsky.app/profile/open...
Open the full discussion →
Primary source reflection.ai ↗
More conversations

What experts are discussing without an anchoring article