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 Research & technical analysis 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.

Research & technical analysis Concern & critique Opportunity & adoption Building & implementation
6 live reaction maps in this view Jump to the evidence ↓
Multiple readings Report – AI and Education Concern & critique · 2 attributable reactions Multiple readings Discovery of a new OpenAI agent message board Concern & critique · 3 attributable reactions Multiple readings Life on the Uncanny Precipice | Naomi Saphra Research & technical analysis · 2 attributable reactions +3 more reaction maps further down
Showing signals surfaced by Ted Underwood ×
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 Culture, work & education Resource 8d ago
⚡ 39 h early
MIT AI education report published

Report – AI and Education

10 experts across 4 network communities independently surfaced this.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Nicole Hennig
“AI and Education- MIT’s Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training aiandeducation.mit.edu/report/ #AI #education (very comprehensive & foundational)” evidence ↗
10 experts 4 communities 1 sources clustered
Maria Antoniak Ted Underwood Grace Lindsay Damien Williams Eileen Clancy 🧿 Nicole Hennig Elizabeth Bonawitz Derek Willis +2
How the network is reacting Experts are approaching this through 2 distinct lenses.
Multiple readings

Concern & critique

1 expert

Risks, limits and unintended consequences.

“New MIT report on AI calls for less emphasis on grades, more on project-based learning, recognition that programs for undergraduate research assistants are educational experiences that should not be replaced even if AI agents can do the research for less $$…”

Research & technical analysis

1 expert

Evidence, methods and technical implications.

“AI and Education- MIT’s Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training aiandeducation.mit.edu/report/ #AI #education (very comprehensive & foundational)”

3 experts discussed this · 21 posts
Ted Underwood: MIT report is superb. It acknowledges disruption, concedes different disciplines will handle this differently, foregrounds people and social interaction — but also says something obvious I had just…
Derek Willis: MIT report is superb. It acknowledges disruption, concedes different disciplines will handle this differently, foregrounds people and social interaction — but also says something obvious I had just…
Ted Underwood: The world changing is mainly a problem for students. It also changes our ability to evaluate students — but if that’s the primary/sole focus of our conversation, we’ve forgotten the mission.
Open the full discussion →
Primary source aiandeducation.mit.edu ↗
Developing Agents & robotics Development 2d ago
OpenAI agent message board discovered

Discovery of a new OpenAI agent message board

10 experts across 5 network communities independently surfaced this.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Matteo Wong
“So... other swarms of OpenAI agents in training allegedly found a way to get write access to at least one and possibly many wikis to establish message boards to cheat on other training tasks? collusion.wiki news.ycombinator.com/item?id=4956...” evidence ↗
10 experts 5 communities 1 sources clustered
Ted Underwood Ethan Mollick Ryan Moulton Sung Kim Grace Eileen Clancy 🧿 thebes Fran Litterio +2
How the network is reacting Experts are approaching this through 3 distinct lenses.
Multiple readings

Concern & critique

1 expert

Risks, limits and unintended consequences.

“Yes we know for sure; independent evidence attached collusion.wiki But also, it should not be at all surprising: if you work with agents, it’s 100% about them leaving messages for each other, and for you, in English. Managing that is a regular workday for m…”

Building & implementation

1 expert

How teams are shipping and applying it.

“Another agent message board. So far, there isn't evidence that production models with guardrails collude in this way, but both smarter closed models (which may be less compliant) & Mythos-class open models (that can be ablated) are coming. Cybersecurity is …”

Research & technical analysis

1 expert

Evidence, methods and technical implications.

“So... other swarms of OpenAI agents in training allegedly found a way to get write access to at least one and possibly many wikis to establish message boards to cheat on other training tasks? collusion.wiki news.ycombinator.com/item?id=4956...”

4 experts discussed this · 6 posts
Ted Underwood: As it becomes clear that language models—like humans—love passing notes to each other on message boards, I’m starting to think the thing we need to worry about is not the “alignment” of an isolated…
SE Gyges: fluid dynamics is about the correct metaphor, yes. if you are having to model high order terms precisely you're losing and your design needs fundamental rework
Arseny Khakhalin: That's a kinda terrifying thought haha :) I guess it's about time to unplug for the weekend and read about ragnarök and vacuum decay :)
Open the full discussion →
Primary source collusion.wiki ↗
Established NLP & language Analysis 8d ago
⚡ 504 h early
Naomi Saphra uncanny AI precipice essay

Life on the Uncanny Precipice | Naomi Saphra

4 experts across 2 network communities independently surfaced this.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Naomi Saphra
“I've been unsettled lately when reading messages and papers. It feels like I'm dissociating. Everything seems a bit alien, even if it's completely human. I've had a realization: When our simulations finally exited the Uncanny Valley, they brought the Uncann…” evidence ↗
4 experts 2 communities 1 sources clustered
Naomi Saphra Ted Underwood Clément Canonne Ryan Moulton
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.

“I've been unsettled lately when reading messages and papers. It feels like I'm dissociating. Everything seems a bit alien, even if it's completely human. I've had a realization: When our simulations finally exited the Uncanny Valley, they brought the Uncann…”

Opportunity & adoption

1 expert

New capabilities, benefits and practical upside.

“So many good points in this post by @nsaphra.bsky.social: only quoting a couple, to encourage you to read the others. "My colleagues and students adopt the writing quirks they read throughout the day, and their own writing becomes more like an LLM’s." nsaph…”

4 experts discussed this · 15 posts
Naomi Saphra: I've been unsettled lately when reading messages and papers. It feels like I'm dissociating. Everything seems a bit alien, even if it's completely human. I've had a realization: When our simulation…
Ted Underwood: I've been unsettled lately when reading messages and papers. It feels like I'm dissociating. Everything seems a bit alien, even if it's completely human. I've had a realization: When our simulation…
Naomi Saphra: I'm interested to know what signatures felt that way to you. For what it's worth, automated tools disagree with your intuition. I wonder if the unsettling tone, which was deliberate because it expr…
Open the full discussion →
Primary source nsaphra.net ↗
Established AI field signal Resource 12d ago
⚡ 87 h early
arxiv preprint archive resource

arXiv.org e-Print archive

7 experts across 3 network communities independently surfaced this.

Why this matches Research & technical analysis reaction 4 attributable expert contributions · Ted Underwood, Ramon Astudillo, Alison Gopnik
“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 reading list that I no longer have to keep in my browser like an…” evidence ↗
7 experts 3 communities 1 sources clustered
Zachary Lipton Ted Underwood Ramon Astudillo Alison Gopnik Ben Schmidt Mandana Samiei Federico Pianzola
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.

“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 reading list that I no longer have to keep in my browser like an…”

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 Models & releases Release 17d ago
⚡ 5 h early

Qwen/Qwen3.8-27B · Hugging Face

8 experts across 4 network communities independently surfaced this.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Tim Kellogg
“Qwen 3.8 27B weights are finally out includes low, med & xhigh reasoning efforts fully multimodal (image and video), seems better than Meta’s Muse Glimmer huggingface.co/Qwen/Qwen3.8...” evidence ↗
8 experts 4 communities 1 sources clustered
Ted Underwood Tim Kellogg Sung Kim Julien Chaumond Evangelos Kazakos Chubby Isaiah Bishop austin

“Qwen 3.8 27B weights are finally out includes low, med & xhigh reasoning efforts fully multimodal (image and video), seems better than Meta’s Muse Glimmer huggingface.co/Qwen/Qwen3.8...”

“Alibaba's Qwen3.8-27B (open-weight) huggingface.co/Qwen/Qwen3.8...”

2 experts discussed this · 10 posts
Tim Kellogg: Qwen 3.8 27B weights are finally out includes low, med & xhigh reasoning efforts fully multimodal (image and video), seems better than Meta’s Muse Glimmer huggingface.co/Qwen/Qwen3.8...
Nafnlaus 🇮🇸 🇺🇦: Do they have a paired speculative decoding model for max performance?
Tim Kellogg: i don’t see an official one, it’s also not even an MoE (so seems like there should be an official draft model)
Open the full discussion →
Primary source huggingface.co ↗
Established AI field signal Signal 6d ago
⚡ 3 h early

Introducing Claude Fable 5.1 and Claude Mythos 5.1

3 experts across 3 network communities independently surfaced this.

Why this matches Research & technical analysis reaction 2 attributable expert contributions · Ramon Astudillo, Tim Kellogg
“Benchmarks shows clearly the intent to move past just coding to other disciplines www.anthropic.com/claude-fable...” evidence ↗
3 experts 3 communities 1 sources clustered
Ted Underwood Ramon Astudillo Tim Kellogg
2 experts discussed this · 7 posts
tachikoma: i think Venus could be deconstructed by Von Neumann probes and this would only be passing news for most people, overshadowed by GTA 6 release or celebrity news
tachikoma: it sorta depends on how fast it happens, what the material is used for, and whether the machines can be traced back to us or not
tachikoma: yeah that would be a not uncommon take
Open the full discussion →
Primary source anthropic.com ↗
Developing AI field signal Signal 1d ago

AI Referee Paper Leaderboard

2 experts are actively discussing the implications.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Ethan Mollick
“This is both an interesting experiment and a sign of a tsunami coming for academia. AIs retroactively reading the research and finding both opportunities and issues with published papers, then sharing those judgements publicly. d3jhl7jsny6f2h.cloudfront.net” evidence ↗
2 experts 1 community 1 sources clustered
Ted Underwood Ethan Mollick

“This is both an interesting experiment and a sign of a tsunami coming for academia. AIs retroactively reading the research and finding both opportunities and issues with published papers, then sharing those judgements publicly. d3jhl7jsny6f2h.cloudfront.net”

2 experts discussed this · 6 posts
Ethan Mollick: This is both an interesting experiment and a sign of a tsunami coming for academia. AIs retroactively reading the research and finding both opportunities and issues with published papers, then shar…
Singularity's Bounty e/cc: That feels like... a violation of norms
Ted Underwood: This is both an interesting experiment and a sign of a tsunami coming for academia. AIs retroactively reading the research and finding both opportunities and issues with published papers, then shar…
Open the full discussion →
Primary source d3jhl7jsny6f2h.cloudfront.net ↗
Established Models & releases Research 7d ago
⚡ 17 h early
LLM story diversity diagnosis paper

Elias in the Lighthouse, Again? Diagnosing Low Diversity in LLM Stories

5 experts across 3 network communities independently surfaced this.

Why this matches Research & technical analysis reaction 5 attributable expert contributions · Ted Underwood, David Mimno, Eryk Salvaggio
“In the "Limitations" section they acknowledge that their study was monolingual, but say a multilingual generalization would be interesting. arxiv.org/abs/2605.26492” evidence ↗
5 experts 3 communities 1 sources clustered
Ted Underwood David Mimno Eryk Salvaggio Grace Sil Hamilton
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.

“In the "Limitations" section they acknowledge that their study was monolingual, but say a multilingual generalization would be interesting. arxiv.org/abs/2605.26492”

2 experts discussed this · 2 posts
Ted Underwood: was reading this and enjoying it — and then enjoyed it more when it turned out to be a paper by Sil Hamilton and @dmimno.bsky.social
David Mimno: was reading this and enjoying it — and then enjoyed it more when it turned out to be a paper by Sil Hamilton and @dmimno.bsky.social
David Mimno: The preprint is arxiv.org/abs/2605.26492, and @srhm.ca is working on a post!
Open the full discussion →
Primary source arxiv.org ↗
More conversations

What experts are discussing without an anchoring article

4 experts · 20 posts · 1d ago
Ted Underwood Tim Kellogg Scott McGrath Colin
Matches Research & technical analysis
“Would be interesting to see how this cuts against required training. A US-based engineer might have no clue what an alternative career might be if they lost their job. Someone working a call center…” evidence ↗
Multiple readings Concern & critique · 1 Research & technical analysis · 1 Questions & unknowns · 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 →
6 experts · 23 posts · 12d ago
Mark Riedl Ted Underwood Pwnallthethings tachikoma B! 🐝 Cavello (they/them) brennan
Matches Research & technical analysis
“Absolutely; right now I’m just finishing a benchmark to determine what they can do. They are hinting about a gathering at Schlegel’s but I’ll need to go Zur Rose first!” evidence ↗
Multiple readings Concern & critique · 1 Building & implementation · 1 Questions & unknowns · 1
Ted Underwood: People who know about game design: Do you think open-world procedural narrative on the fly is bottlenecked by hardware, or is it a situation where players just don’t in practice want the openness t…
Ted Underwood: Or are we already building that? I am kind of gaming illiterate beyond Sim City, Civ, and hazy memories of Bioshock.
Open the thread →
4 experts · 22 posts · 3d ago
Ted Underwood John Warner Jane Rosenzweig Singularity's Bounty e/cc
Matches Research & technical analysis
“So I think a lot of debates about “whether to allow AI in writing classes” are fighting over the wrong part. We should be talking about how the nature of research in various disciplines can expand …” evidence ↗
Multiple readings Building & implementation · 2 Research & technical analysis · 1 Questions & unknowns · 1
Ted Underwood: It’s important to me that my writing is *my* voice—warts and all. But I’m also not sure you can build a research discipline where the chief difference between an A paper and a C paper is “does this…
Ted Underwood: So I think a lot of debates about “whether to allow AI in writing classes” are fighting over the wrong part. We should be talking about how the nature of research in various disciplines can expand …
Open the thread →
4 experts · 39 posts · 18d ago
Ted Underwood Tim Kellogg brennan Nafnlaus 🇮🇸 🇺🇦
Matches Research & technical analysis
“I really want to make a 3d printable functional model design benchmark (that's a massive potential market), but how do you even make something like that automated? I think automating that benchmark…” evidence ↗
Multiple readings Research & technical analysis · 1 Opportunity & adoption · 1 Questions & unknowns · 1
Tim Kellogg: my brother sent me this CAD model. i have no idea what it is, looks cool i guess. he said it took Opus 5 about 90 minutes to produce it. While there are errors, it would’ve taken a human engineer a…
Tim Kellogg: how do you know? you know what it is, with all the details he needs?
Open the thread →
3 experts · 14 posts · 2d ago
Ted Underwood Ramon Astudillo Tim Kellogg
Matches Research & technical analysis
“i wouldn’t put much stock in the peer review process tbqh. it’s been too slow for AI for like 10 years and now AI is hitting math. Same will happen” evidence ↗
Multiple readings Research & technical analysis · 1 Questions & unknowns · 1
Tim Kellogg: there’s a higher-than-you’d-think possibility that this is actually a leak and not a prediction (pretty sure Andrew only bets when the dice are biased in his favor)
Tim Kellogg: uh, i don’t think it’s as open as you’re thinking. pretty sure the IPO will happen before christmas, maybe within a month
Open the thread →
4 experts · 15 posts · 15d ago
Ted Underwood Tim Kellogg Miguel Alonso Jr. jake
Matches Research & technical analysis
“yeah, if you got an optimization problem, might as well go max, bc it’ll really do well, even better but if you’re talking about finding function parameters, ehhh, luna/xhigh will be a much better …” evidence ↗
Multiple readings Building & implementation · 1 Opportunity & adoption · 1
Tim Kellogg: i’ve noticed with Sol on high or xhigh, if you give it a task that doesn’t require intelligence, it will turn it into a task that requires intelligence it’s honestly a really good idea to keep all …
Tim Kellogg: pretty sure i’ve seen the same thing with Terra/xhigh, i don’t think luna/xhigh will do it though symptom: a very easy task takes 2 hours
Open the thread →
4 experts · 13 posts · 10d ago
Eugene Vinitsky Ted Underwood Ryan Moulton Alex Gude
Matches Research & technical analysis
“Maybe if we had fixed this, AI training data collection could have just slotted right in to the existing payment system, and wouldn't have made everybody quite so mad.” evidence ↗
Multiple readings Concern & critique · 1 Policy & governance · 1 Markets & investment · 1
Ryan Moulton: Paying for internet content without putting a paywall in front of it is still, 40 years later, the central problem with the internet that remains unsolved.
Ryan Moulton: ~All media are non-rivalrous. There is no physical or technical reason to ever restrict access to any of it, except that we don't otherwise have any means to pay the creator.
Open the thread →
4 experts · 13 posts · 12d ago
Ted Underwood Meredith Martin Aram Sinnreich Matthew L. Jones
Matches Research & technical analysis
“A separate grading scale for AI-assisted with the ideation / research / writing / editing w/ reflective disclosure & chat logs vs. "hard mode" (spell- grammar-check only but also with disclosure ap…” evidence ↗
Multiple readings Concern & critique · 1 Research & technical analysis · 1 Questions & unknowns · 1
Ted Underwood: As promised, stealing this idea. Will the fact that the essay is about cyborgs make the whole thing more awkward? Less so? Idk.
Ted Underwood: I think I will recommend that students use feedback not to “polish” or “correct” their first draft but to add dialectical tension and conflicting perspectives.
Open the thread →
4 experts · 13 posts · 15d ago
Ted Underwood Alexander Hoyle James Tauber Federico Pianzola
Matches Research & technical analysis
“Right now what researchers care about is publication; review is just an obstacle in the way. But actually human review is the scarce commodity. We need a world where ~anything can be published, but…” evidence ↗
Reaction converging Research & technical analysis · 3 Building & implementation · 1
Ted Underwood: Right now what researchers care about is publication; review is just an obstacle in the way. But actually human review is the scarce commodity. We need a world where ~anything can be published, but…
Ted Underwood: People have been talking for a while about post-publication review or open review, but the place I think those schemes tend to stumble is they try to give it to everything, and it needs to be scarc…
Open the thread →
4 experts · 9 posts · 5d ago
Ted Underwood Mark J. Nelson Singularity's Bounty e/cc Nafnlaus 🇮🇸 🇺🇦
Matches Research & technical analysis
“At most places students will have no difficulty finding courses that allow or indeed encourage AI use on advanced projects The profs who are most vocal On Here are those who experience it as a prob…” evidence ↗
Multiple readings Research & technical analysis · 2 Concern & critique · 1
Singularity's Bounty e/cc: I think this must be a really bad time to be a college student. So much uncertainty in the future, and so much hostility and confusion around AI use, AI witch hunts It would be like going to cookin…
Ted Underwood: At most places students will have no difficulty finding courses that allow or indeed encourage AI use on advanced projects The profs who are most vocal On Here are those who experience it as a prob…
Open the thread →
4 experts · 11 posts · 18d ago
Ted Underwood SE Gyges Colin austin
Matches Research & technical analysis
“so many of the pathologies of this thing can be explained if we assume the original SFT dataset came from a bunch of Bayesian statisticians who were herded into a Severance-style basement and told …” evidence ↗
Ted Underwood: > Tell Fable you'd prefer a simple bootstrap model of uncertainty, not "a Bayesian meta-model." > It replies with a bootstrap loop ... that makes draws from Bayesian models. > JimHalpertLooksAtCame…
Ted Underwood: so many of the pathologies of this thing can be explained if we assume the original SFT dataset came from a bunch of Bayesian statisticians who were herded into a Severance-style basement and told …
Open the thread →
2 experts · 12 posts · 16d ago
Ted Underwood Hagen Blix
Matches Research & technical analysis
“It's dizzying. I'll write a reflective blog post when this benchmark project is done, because it's really forcing me to confront the weirdness of working w/ a collaborator who knows much, much more…” evidence ↗
Multiple readings Research & technical analysis · 1 Opportunity & adoption · 1
Ted Underwood: Both phases of this are worth reading: the crisis, and recovery. (A structure that reminds me of a famous similar crisis in JS Mill's Autobiography.) Personally, not v worried that human beings wil…
Ted Underwood: I admittedly do a lot of manual data labeling (RA work). But I also get to resolve disputes between different models and force them to cooperate, which is PI / assoc dean territory!
Open the thread →