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
1 live reaction map in this view Jump to the evidence ↓
Shared emphasis Pretraining Recurrent Networks without Recurrence Research & technical analysis · 2 attributable reactions
Showing signals surfaced by Evangelos Kazakos ×
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.

Established Models & releases Release 16d 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 NLP & language Research 12d ago
⚡ 3 h early

Pretraining Recurrent Networks without Recurrence

4 experts across 3 network communities independently surfaced this.

Why this matches Research & technical analysis reaction 2 attributable expert contributions · Christopher Manning, Sung Kim
“Paper: https://t.co/IoDcRINaAL This is one of a whole bunch of recent papers reviving study of recurrent neural networks. One weird omission is not testing LSTM RNNs. Surely they remain the canonical successful RNN architecture? Another completely uninvesti…” evidence ↗
4 experts 3 communities 1 sources clustered
Christopher Manning Ted Underwood Sung Kim Evangelos Kazakos
How the network is reacting 2 experts are emphasizing research & technical analysis.
Shared emphasis

Research & technical analysis

2 experts

Evidence, methods and technical implications.

“Pretraining Recurrent Networks without Recurrence It sidesteps the limitations of RNNs by using a Transformer teacher to learn strong predictive state representations, then uses supervised learning to train the memory transition function. arxiv.org/abs/2606…”

Primary source arxiv.org ↗
Established Models & releases Research 5d ago
⚡ 2210 h early
VLM suppresses female representation paper

Vision-Language Models Suppress Female Representations Under Ambiguous Input

2 directory members surfaced this signal.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · AI Firehose
“Research shows vision-language models (VLMs) default to male in ambiguous images, outputting male for female jobs. While alignment improves outputs, it masks encoded biases, raising fairness issues in AI applications. https://arxiv.org/abs/2605.31556” evidence ↗
2 experts 2 communities 1 sources clustered
Evangelos Kazakos AI Firehose

“Research shows vision-language models (VLMs) default to male in ambiguous images, outputting male for female jobs. While alignment improves outputs, it masks encoded biases, raising fairness issues in AI applications. https://arxiv.org/abs/2605.31556”

Primary source arxiv.org ↗
Developing AI research Research 1d ago
⚡ 24 h early
AutoCompass visual localization paper

AutoCompass: Accurate Visual Localization on Public Maps by Learning from Weak Labels

2 directory members surfaced this signal.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Eric Brachmann
“New #ECCV2026 paper alert: 🧭 AutoCompass: Accurate Visual Localization on Public Maps by Learning from Weak Labels 🧭 Paper: arxiv.org/abs/2609.02798 Website: nianticspatial.github.io/autocompass/ Video: www.youtube.com/watch?v=hmUF...” evidence ↗
2 experts 1 community 3 sources clustered
Eric Brachmann Evangelos Kazakos

“New #ECCV2026 paper alert: 🧭 AutoCompass: Accurate Visual Localization on Public Maps by Learning from Weak Labels 🧭 Paper: arxiv.org/abs/2609.02798 Website: nianticspatial.github.io/autocompass/ Video: www.youtube.com/watch?v=hmUF...”

Primary source arxiv.org ↗ nianticspatial.github.io ↗ youtube.com ↗
Established AI field signal Signal 17d ago
⚡ 4 h early

biglam/british-library-book-images · Datasets at Hugging Face

2 experts are actively discussing the implications.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Daniel van Strien
“Dataset (crop_masks config): huggingface.co/datasets/big... Search + cutouts: huggingface.co/spaces/davan... Model: huggingface.co/small-models...” evidence ↗
2 experts 2 communities 1 sources clustered
Daniel van Strien Evangelos Kazakos

“Dataset (crop_masks config): huggingface.co/datasets/big... Search + cutouts: huggingface.co/spaces/davan... Model: huggingface.co/small-models...”

2 experts discussed this · 2 posts
Daniel van Strien: Uploaded a dataset of 1,080,814 public domain images, mostly from 19th-century books, to the Hugging Face Hub: huggingface.co/datasets/big... You can also do semantic search against the images here…
Evangelos Kazakos: Uploaded a dataset of 1,080,814 public domain images, mostly from 19th-century books, to the Hugging Face Hub: huggingface.co/datasets/big... You can also do semantic search against the images here…
Lynn Cherny: ah, can you dm me?
Open the full discussion →
Primary source huggingface.co ↗
Established AI field signal Signal 17d ago
⚡ 4 h early

British Library Book Image Search - a Hugging Face Space by davanstrien

2 directory members surfaced this signal.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Daniel van Strien
“Dataset (crop_masks config): huggingface.co/datasets/big... Search + cutouts: huggingface.co/spaces/davan... Model: huggingface.co/small-models...” evidence ↗
2 experts 2 communities 1 sources clustered
Daniel van Strien Evangelos Kazakos

“Dataset (crop_masks config): huggingface.co/datasets/big... Search + cutouts: huggingface.co/spaces/davan... Model: huggingface.co/small-models...”

Primary source huggingface.co ↗
Established NLP & language Signal 17d ago
⚡ 19 h early

The IOL-AI Challenge: An Open Challenge towards Advancing Linguistic Reasoning

2 directory members surfaced this signal.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Cohere Labs
“@UvA_Amsterdam @TheyCallMeMr_ @johnamqdang @mgalle @mziizm @ahmetustun89 @beyzaermis @ananya__sahu @mhrnz_m @mrdanieldsouza @rsrchvslstrytlr @ConorFinlay3 @joshuakurien_ Announcing new research on the state of the art of linguistic reasoning. 🧩 Led by @DanM…” evidence ↗
2 experts 2 communities 1 sources clustered
Cohere Labs Evangelos Kazakos

“@UvA_Amsterdam @TheyCallMeMr_ @johnamqdang @mgalle @mziizm @ahmetustun89 @beyzaermis @ananya__sahu @mhrnz_m @mrdanieldsouza @rsrchvslstrytlr @ConorFinlay3 @joshuakurien_ Announcing new research on the state of the art of linguistic reasoning. 🧩 Led by @DanM…”

Primary source arxiv.org ↗
Established Culture, work & education Signal 10d ago

Ai2 and Providence Swedish Cancer Institute partner to advance AI-assisted scientific discovery | Ai2

2 directory members surfaced this signal.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Ai2
“AI’s biggest role in science may not be answering questions. It may be helping scientists find which questions are worth asking. At Providence Swedish, AutoDiscovery surfaced an unexpected immune signal in a heavily studied cancer dataset—and follow-up rese…” evidence ↗
2 experts 2 communities 1 sources clustered
Ai2 Evangelos Kazakos

“AI’s biggest role in science may not be answering questions. It may be helping scientists find which questions are worth asking. At Providence Swedish, AutoDiscovery surfaced an unexpected immune signal in a heavily studied cancer dataset—and follow-up rese…”

Primary source buff.ly ↗
Established AI field signal Signal 18d ago

When a model reads a drug's class from its name—not its knowledge | Ai2

2 directory members surfaced this signal.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · Ai2
“A model can sound like it knows a drug—even when it doesn’t. Researchers at @utaustin.bsky.social, @northeasternu.bsky.social, & @mdanderson.bsky.social found LLMs often lack drug-specific knowledge and instead lean on morphology. They used Olmo to trace wh…” evidence ↗
2 experts 2 communities 1 sources clustered
Ai2 Evangelos Kazakos

“A model can sound like it knows a drug—even when it doesn’t. Researchers at @utaustin.bsky.social, @northeasternu.bsky.social, & @mdanderson.bsky.social found LLMs often lack drug-specific knowledge and instead lean on morphology. They used Olmo to trace wh…”

Primary source buff.ly ↗
Established AI research Resource 20d ago
Oslo ML cognitive neuroscience PhD position

PhD Research Fellow in Machine Learning for Cognitive Neuroscience (306376) | University of Oslo

2 directory members surfaced this signal.

Why this matches Research & technical analysis reaction 1 attributable expert contribution · ELLIS
“🏹 Job Alert: PhD Research Fellow in Machine Learning for Cognitive Neuroscience 📍 Oslo 🇳🇴 ⏰ 16 September 🔗 https://www.jobbnorge.no/en/available-jobs/job/306376/phd-research-fellow-in-machine-learning-for-cognitive-neuroscience” evidence ↗
2 experts 2 communities 1 sources clustered
ELLIS Evangelos Kazakos

“🏹 Job Alert: PhD Research Fellow in Machine Learning for Cognitive Neuroscience 📍 Oslo 🇳🇴 ⏰ 16 September 🔗 https://www.jobbnorge.no/en/available-jobs/job/306376/phd-research-fellow-in-machine-learning-for-cognitive-neuroscience”

Primary source jobbnorge.no ↗
Debates gaining attention

What experts are discussing without an anchoring article

2 experts · 3 posts · 2d ago
Evangelos Kazakos jvgemert.bsky.social
Matches Research & technical analysis
“This is why I work on data efficiency: to allow AI training for the long tail of SMEs and universities who don't have the huge resources. I don't believe AI will go away. Let's make training more a…” evidence ↗
Multiple readings Research & technical analysis · 1 Opportunity & adoption · 1
Evangelos Kazakos: Given everything that is going on, I’m more demotivated than ever to contribute to ‘AI’, whatever that term means. And no, currently I see literally zero benefits to society. AI for good is the sca…
jvgemert.bsky.social: This is why I work on data efficiency: to allow AI training for the long tail of SMEs and universities who don't have the huge resources. I don't believe AI will go away. Let's make training more a…
Open the thread →