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 Opportunity & adoption 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.

Opportunity & adoption
0 live reaction maps in this view

No development in this filter has enough independent reactions yet. The maps appear as soon as the evidence supports them.

Most important

The developments commanding sustained expert attention

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

New Models & releases Release 17h ago
⚡ 14 h early
Ripple 65B preview model release

ricursiveAI/Ripple-65B-Preview · Hugging Face

2 directory members surfaced this signal.

Why this matches Opportunity & adoption reaction 1 attributable expert contribution · Ramon Astudillo
“> modernized dense Transformer model rather than [MoE] > thin-and-deep model aspect ratio, [..] d_model (5120) and 80 total layers. We switch from [..] llama-like to a Gemma 3N-like huggingface.co/ricursiveAI/Ripple-65B...” evidence ↗
2 experts 2 communities 1 sources clustered
Ramon Astudillo Sung Kim

“> modernized dense Transformer model rather than [MoE] > thin-and-deep model aspect ratio, [..] d_model (5120) and 80 total layers. We switch from [..] llama-like to a Gemma 3N-like huggingface.co/ricursiveAI/Ripple-65B...”

“Ripple-65B Preview, an open-weight model on agentic coding. huggingface.co/ricursiveAI/...”

Primary source huggingface.co ↗
New Culture, work & education Development 15h ago
⚡ 5 h early
Google AI overviews forced in classrooms

Google Is Forcing AI Overviews in Classrooms. Schools Are Searching for a Way Out

2 experts are actively discussing the implications.

Why this matches Opportunity & adoption reaction 1 attributable expert contribution · Ben Williamson
“"This seems counterproductive to what we're trying to teach.... Students need opportunities to navigate search results, examine sources, and draw their own conclusions rather than being handed anAI-generated synthesis before they've even started.” www.wired…” evidence ↗
2 experts 2 communities 2 sources clustered
Ben Williamson Danya Glabau

“"This seems counterproductive to what we're trying to teach.... Students need opportunities to navigate search results, examine sources, and draw their own conclusions rather than being handed anAI-generated synthesis before they've even started.” www.wired…”

2 experts discussed this · 10 posts
Ben Williamson: "... it should not be down to tech giants to set education policy" - great to see this in New Scientist in response to Google's efforts to make AI the default medium for knowledge access in schools…
Ben Williamson: Google has been a problematic policy actor in education for years, far pre-dating Gemini. Its education platform Classroom was already a vector for automation in education www.edweek.org/technology…
Ben Williamson: Before ChatGPT, Google had already made its plans to introduce LLMs into education well known. In 5 years it's gone from shonky demos to forcing it into schools via existing Chromebook purchases an…
Open the full discussion →
Primary source wired.com ↗ newscientist.com ↗
New AI field signal Signal 21h ago

Ai exponential time hypothesis

1 directory member surfaced this signal.

Why this matches Opportunity & adoption reaction 1 attributable expert contribution · Simons Institute for the Theory of Computing
“@nytimes.com story out today on the OpenAI breakthrough on the Exponential Time Hypothesis, featuring our friends Russell Impagliazzo and @rrwilliams.bsky.social. www.nytimes.com/2026/11/06/s... .” evidence ↗
1 expert 1 community 1 sources clustered
Simons Institute for the Theory of Computing

“@nytimes.com story out today on the OpenAI breakthrough on the Exponential Time Hypothesis, featuring our friends Russell Impagliazzo and @rrwilliams.bsky.social. www.nytimes.com/2026/11/06/s... .”

Primary source nytimes.com ↗
New Vision & synthetic media Research 10h ago
scale-invariant diffusion continuous super-resolution

Everything at Every Scale: Scale-Invariant Diffusion with Continuous Super-Resolution

1 directory member surfaced this signal.

Why this matches Opportunity & adoption reaction 1 attributable expert contribution · AI Firehose
“The SKILD model combines generation and super-resolution using scale-invariant diffusion for smooth task transitions. This efficient design achieves leading image quality on CIFAR-10 and strong super-resolution on ImageNet, potentially transforming image pr…” evidence ↗
1 expert 1 community 1 sources clustered
AI Firehose

“The SKILD model combines generation and super-resolution using scale-invariant diffusion for smooth task transitions. This efficient design achieves leading image quality on CIFAR-10 and strong super-resolution on ImageNet, potentially transforming image pr…”

Primary source arxiv.org ↗
Developing Agents & robotics Signal 1d ago

agents.md

1 directory member surfaced this signal.

Why this matches Opportunity & adoption reaction 1 attributable expert contribution · trey.io
“I'm my global agents.md & claude.md files I tell Claude to always use a uv script with a shebang line. It's been great for skills that involve scripts with third party dependencies. Very fast and no risk of cluttering up the global Python environments.” evidence ↗
1 expert 0 communities 1 sources clustered
trey.io

“I'm my global agents.md & claude.md files I tell Claude to always use a uv script with a shebang line. It's been great for skills that involve scripts with third party dependencies. Very fast and no risk of cluttering up the global Python environments.”

Primary source agents.md ↗
Debates gaining attention

What experts are discussing without an anchoring article

5 experts · 13 posts · 22h ago
Ted Underwood Andrew Perfors Andrew Piper Meredith Martin D. Hicks
Matches Opportunity & adoption
“oh, they already do, but I suspect my agent can say "no thank you" faster” evidence ↗
Multiple readings Concern & critique · 1 Research & technical analysis · 1 Context & explanation · 1
Ted Underwood: We need a social movement to explain to students and entrepreneurs — that it's a bad idea to email random professors in other states for advice. It was charming to get one inquiry a month, but thre…
Ted Underwood: I'm not sure what happened; GenAI is part of it, but not all of it. The trend began earlier.
Open the thread →
3 experts · 22 posts · 14h ago
Blake Richards Dan Goodman Cian O'Donnell
Matches Opportunity & adoption
“I was mostly thinking about the brain in the post at the top, not ML. I can certainly see that comparing expectations to observations is useful, but isn't that obvious? That's not the same as repla…” evidence ↗
Shared emphasis Questions & unknowns · 2
Dan Goodman: Genuine question asked out of ignorance and not to make a point: is "replay" still a useful concept? I'm not an expert on this at all, but it seems like you would have had to have done all the inte…
Dan Goodman: I don't really understand this.
Open the thread →
3 experts · 3 posts · 11h ago
Mark J. Nelson James Tauber Al Merose
Matches Opportunity & adoption
“My students weren't very happy to discover that in machine learning, "tensor" is just a puffed-up jargon term for a multi-dimensional array.” evidence ↗
Mark J. Nelson: My students weren't very happy to discover that in machine learning, "tensor" is just a puffed-up jargon term for a multi-dimensional array.
Al Merose: We can all blame Jeff Dean for that.
Open the thread →
2 experts · 6 posts · 1d ago
SE Gyges Colin
Matches Opportunity & adoption
“It’s not really that, exactly. The math thing is just the LLM doing some data analysis and then tweaking the model, much in the exact same way that a skilled human might. It’s impressive but not mo…” evidence ↗
SE Gyges: this is an insanely funny admission given that the post is still correct in terms of all of the literal words that are literally in the post
Colin: there's just nothing really to address. The math performance thing on the small Qwen model is just a programming task. A kind of strange one, but not an especially impossible one. It doesn't reveal…
Open the thread →