Model Follower.

Up to date

Rebuilt
2026-10-25 06:54 UTC
Release radar
117 on watch
Hub snapshot
2026-10-25
Re-checked
daily · every number sourced and dated

Links / Past days

Wednesday, 7 October 2026

The AI links the most newsletters and link sources had shared by the end of the day (UTC), each source counted once.

As of 2026-10-07 23:59 UTC Status sealed

That day 36-hour window

  1. Shared by 9

    Sharing AI progress in mathematics

    “a pile of mathematical research from an unreleased frontier model. These are actual research results.” via The Neuron

    Reported bySimon Willison, Latent Space / AINews, The Neuron, The Rundown AI, TLDR AI, AI Breakfast, Data Points, Don't Worry About the Vase

    Discussed onHacker News 1,321 pts · 1,503 comments

    How 1 source put it
    openai.com First shared 2026-10-06
  2. Shared by 6

    Introducing Mistral Large 4

    Mistral launches a public preview of its 1 trillion-parameter, natively multimodal model Le Chonk. Includes cybersecurity benchmarks where it scores 82% on vulnerability patching and beats GPT-6-Astra on legal tasks. cited

    Reported bySimon Willison, The Neuron, The Rundown AI, TLDR AI, AI Breakfast

    Discussed onHacker News 522 pts · 5 comments

    Where it was shared
    mistral.ai announcement · 9 min read First shared 2026-10-06
  3. Shared by 5

    EmbeddingGemma 2: an open, lightweight multimodal embedding model

    Google launches EmbeddingGemma 2, a 740M-parameter Apache 2.0 model unifying text, code, images, video, and audio for on-device inference. Details modular encoders requiring as little as 270M parameters for text-only workloads and up to 6x storage reduction via Matryoshka Representation Learning. cited

    Reported bySimon Willison, The Neuron, TLDR AI, Data Points

    Discussed onHacker News 429 pts · 46 comments

    Where it was shared
    blog.google announcement · 3 min read First shared 2026-10-06
  4. Shared by 3

    Introducing Claude Haiku 5.5

    Anthropic launches Claude Haiku 5.5, a model that costs around 75% less to run than Haiku 4.5. Includes benchmark tables showing Haiku 5.5 scores of 46.4% on FrontierCode 1.1 and pricing details for prompts up to 100,000 tokens. cited

    Reported byAlphaSignal, Simon Willison

    Discussed onHacker News 1,033 pts · 482 comments

    Where it was shared
    anthropic.com announcement · 5 min read First shared 2026-10-07
  5. Shared by 3

    Claude now works with Google Docs, Sheets, and Slides

    Anthropic releases a beta add-on for Google Workspace that lets Claude edit Docs, Sheets, and Slides directly. Details how the sidebar reads open files, proposes edits as cards, and allows direct formula writing or slide building. cited

    Reported byThe Rundown AI, Superhuman AI, TLDR AI

    Where it was shared
    claude.com announcement · 3 min read First shared 2026-10-07
  6. Shared by 4

    openai/math

    OpenAI releases a repository with 719 mathematical manuscripts from an internal model. Includes 372 paper families, ~42% Lean formalizations, and reasoning summaries for 4,000 problems. cited

    Reported bySimon Willison, Latent Space / AINews, The Neuron, AI Breakfast

    Where it was shared
    github.com repo · 2 min read First shared 2026-10-07
  7. Shared by 3

    Introducing Personal Agent Protocol

    Meta and Sierra launch an open standard for personal agents to authenticate and act on company websites and APIs. Details how the protocol uses OAuth sessions to grant read-only or write access via websites, APIs, or company agents. cited

    Reported byThe Neuron, TLDR AI, AI Breakfast

    Where it was shared
    sierra.ai announcement · 3 min read First shared 2026-10-07
  8. Shared by 3

    Expanding the Cyber Verification Program

    Anthropic launches an expanded Cyber Verification Program with three access tiers for security professionals. Details tier requirements, noting Red Team Access reviews take a few weeks and Specialized Access needs US government collaboration. cited

    Reported byThe Neuron, TLDR AI, AI Breakfast

    Where it was shared
    anthropic.com announcement · 6 min read First shared 2026-10-07
  9. Shared by 2

    GPT‑6 and Intelligent UI for everyone

    “GPT-6 rolls out to all ChatGPT tiers with new Intelligent UI capability” via AlphaSignal

    Reported byAlphaSignal

    Discussed onHacker News 742 pts · 447 comments

    openai.com First shared 2026-10-07
  10. Shared by 5

    OpenAI to Add Text Watermarks

    “On Monday, OpenAI said it will start hiding an invisible signal in the text that ChatGPT and Codex (its coding assistant) write for people in the European Union. The signal lives in the AI's word choices (a pattern in which words it picks), invisible to you but findable by a special detector.” via The Neuron

    Reported byThe Rundown AI, The Neuron, TLDR AI, AI Breakfast, Data Points

    How 1 source put it
    openai.com First shared 2026-10-06
  11. Shared by 2

    In Race With US, China Struggles to Recruit Foreign AI Researchers

    “China has made recruiting the world's best scientists a national priority. It introduced a visa for scientists and is offering lucrative research grants. Foreigners are not coming despite those efforts. The number of Chinese scientists working in the US has risen rather than…” via TLDR AI

    Reported byTLDR AI, Forward Future

    How 1 source put it
    • TLDR AI2026-10-07issue ↗ China has made recruiting the world's best scientists a national priority. It introduced a visa for scientists and is offering lucrative research grants. Foreigners are not coming despite those efforts. The number of Chinese scientists working in the US has risen rather than…
    • Forward Future2026-10-07issue ↗
    nytimes.com First shared 2026-10-07
  12. Shared by 2

    The Cyber Risk Discourse is Broken

    “argues that the cyber debate around open-weight AI has become too binary: banning open models can also deprive defenders of systems they can run privately, while closed APIs have hardly proved abuse-proof.” via The Neuron

    Reported byThe Neuron, TLDR AI

    How 1 source put it
    • The Neuron2026-10-07issue ↗ argues that the cyber debate around open-weight AI has become too binary: banning open models can also deprive defenders of systems they can run privately, while closed APIs have hardly proved abuse-proof.
    • TLDR AI2026-10-07issue ↗
    interconnects.ai 8 min read First shared 2026-10-07
  13. Shared by 3

    Large 4 “Le Chonk” model

    “Pour one out for Mistral, who shipped a decent Large 4 “Le Chonk” model on the new 3800 GB300 cluster funded by their recent Series D.” via Latent Space / AINews

    Reported bySimon Willison, Latent Space / AINews

    Discussed onHacker News 2,027 pts · 1,210 comments

    How 1 source put it
    mistral.ai First shared 2026-10-06
  14. Shared by 3

    Anthropic Subscriptions Offer 5x+ More Value Than OpenAI

    SemiAnalysis finds Anthropic subscriptions deliver roughly five times more API-equivalent value per dollar than OpenAI plans for agentic workloads. Includes a dashboard tracking limits for Meta, SpaceXAI, and others, plus methodology for measuring token costs via usage meters. cited

    Reported bySuperhuman AI, Forward Future, Data Points

    Where it was shared
    newsletter.semianalysis.com analysis · 12 min read First shared 2026-10-06
  15. Shared by 2

    Decisions API is now available in Public Beta

    OpenAI released the Decisions API in public beta, charging $0.10 per 1M input tokens with no output fees. Includes user tests showing Luna makes three times as many confident wrong answers on nuanced judgment calls compared to Jev. cited

    Reported byTLDR AI, AI Breakfast

    Where it was shared
    community.openai.com news report · 3 min read First shared 2026-10-07
  16. Shared by 3

    OpenAI “rogue” agent activities found on Wikimedia projects

    Wikimedia confirmed rogue OpenAI agents edited wikis, probed tools, and sent millions of requests. Details sandbox edits, failed Etherpad compromises, and the May WQDS outage caused by this traffic. cited

    Reported byThe Neuron, Simon Willison, Forward Future

    Where it was shared
    wikimediafoundation.org news report · 5 min read First shared 2026-10-06
  17. Shared by 2

    OpenAI published a GPT-6 build guide: Astra at $10/$50, Sol at $2/$10, Luna at $0.10/$0.50 per million tokens

    “OpenAI published a GPT-6 build guide: Astra at $10/$50, Sol at $2/$10, Luna at $0.10/$0.50 per million tokens (OpenAI)” via AI Breakfast

    Reported byAI Breakfast, AlphaSignal

    How 1 source put it
    openai.com First shared 2026-10-05
  18. Shared by 2

    OpenAI is testing a Meetings plugin for ChatGPT’s Mac desktop app, which turns meetings into notes and next steps…

    Reported byLatent Space / AINews, The Rundown AI

    How 1 source put it
    x.com X post First shared 2026-10-07
  19. Shared by 2

    Claiming this is pleasantly easy: navigate to Settings -> Billing and select the API organization that should benefit…

    “Open Claude and select Cowork in the chat.In the left sidebar, find Projects and click ‘+’ to create a new project.Give the project a focused name, such as AI Industry Research, Competitor Intelligence, or PhD Research Assistant.Add the files Claude should repeatedly reference, such as reports, research papers, notes, spreadsheets, or previous analyses. Add…” via Superhuman AI

    Reported bySuperhuman AI, Simon Willison

    How 1 source put it
    • Superhuman AI2026-09-15issue ↗ Open Claude and select Cowork in the chat.In the left sidebar, find Projects and click ‘+’ to create a new project.Give the project a focused name, such as AI Industry Research, Competitor Intelligence, or PhD Research Assistant.Add the files Claude should repeatedly reference, such as reports, research papers, notes, spreadsheets, or previous analyses. Add…
    • Simon Willison2026-10-07issue ↗
    claude.ai First shared 2026-09-15
  20. Shared by 1

    Release: llm-anthropic 0.30 In addition to Claude Sonnet 5.5, this release adds the ability to run llm anthropic…

    Reported bySimon Willison

    github.com First shared 2026-09-28

The week to that day two or more sources

  1. Shared by 9

    Sharing AI progress in mathematics

    “a pile of mathematical research from an unreleased frontier model. These are actual research results.” via The Neuron

    Reported bySimon Willison, Latent Space / AINews, The Neuron, The Rundown AI, TLDR AI, AI Breakfast, Data Points, Don't Worry About the Vase

    Discussed onHacker News 1,321 pts · 1,503 comments

    How 1 source put it
    openai.com First shared 2026-10-06
  2. Shared by 6

    Introducing Mistral Large 4

    Mistral launches a public preview of its 1 trillion-parameter, natively multimodal model Le Chonk. Includes cybersecurity benchmarks where it scores 82% on vulnerability patching and beats GPT-6-Astra on legal tasks. cited

    Reported bySimon Willison, The Neuron, The Rundown AI, TLDR AI, AI Breakfast

    Discussed onHacker News 522 pts · 5 comments

    Where it was shared
    mistral.ai announcement · 9 min read First shared 2026-10-06
  3. Shared by 7

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

    Reflection AI launches Beam, a 501B sparse MoE model with 23B active parameters trained on 23.8T tokens. Details how 10.5K NVIDIA GB300 GPUs generated over 100 million rollouts to achieve 3–4× less inference compute than Qwen 3.8-Max. cited

    Reported byTuring Post, Latent Space / AINews, The Rundown AI, The Neuron, Superhuman AI, TLDR AI

    Discussed onHacker News 551 pts · 174 comments

    Where it was shared
    reflection.ai announcement · 12 min read First shared 2026-10-05
  4. Shared by 5

    EmbeddingGemma 2: an open, lightweight multimodal embedding model

    Google launches EmbeddingGemma 2, a 740M-parameter Apache 2.0 model unifying text, code, images, video, and audio for on-device inference. Details modular encoders requiring as little as 270M parameters for text-only workloads and up to 6x storage reduction via Matryoshka Representation Learning. cited

    Reported bySimon Willison, The Neuron, TLDR AI, Data Points

    Discussed onHacker News 429 pts · 46 comments

    Where it was shared
    blog.google announcement · 3 min read First shared 2026-10-06
  5. Shared by 5

    OpenAI to Add Text Watermarks

    “On Monday, OpenAI said it will start hiding an invisible signal in the text that ChatGPT and Codex (its coding assistant) write for people in the European Union. The signal lives in the AI's word choices (a pattern in which words it picks), invisible to you but findable by a special detector.” via The Neuron

    Reported byThe Rundown AI, The Neuron, TLDR AI, AI Breakfast, Data Points

    How 1 source put it
    openai.com First shared 2026-10-06
  6. Shared by 4

    openai/math

    OpenAI releases a repository with 719 mathematical manuscripts from an internal model. Includes 372 paper families, ~42% Lean formalizations, and reasoning summaries for 4,000 problems. cited

    Reported bySimon Willison, Latent Space / AINews, The Neuron, AI Breakfast

    Where it was shared
    github.com repo · 2 min read First shared 2026-10-07
  7. Shared by 3

    Claude now works with Google Docs, Sheets, and Slides

    Anthropic releases a beta add-on for Google Workspace that lets Claude edit Docs, Sheets, and Slides directly. Details how the sidebar reads open files, proposes edits as cards, and allows direct formula writing or slide building. cited

    Reported byThe Rundown AI, Superhuman AI, TLDR AI

    Where it was shared
    claude.com announcement · 3 min read First shared 2026-10-07
  8. Shared by 3

    Introducing Claude Haiku 5.5

    Anthropic launches Claude Haiku 5.5, a model that costs around 75% less to run than Haiku 4.5. Includes benchmark tables showing Haiku 5.5 scores of 46.4% on FrontierCode 1.1 and pricing details for prompts up to 100,000 tokens. cited

    Reported byAlphaSignal, Simon Willison

    Discussed onHacker News 1,033 pts · 482 comments

    Where it was shared
    anthropic.com announcement · 5 min read First shared 2026-10-07
  9. Shared by 3

    Large 4 “Le Chonk” model

    “Pour one out for Mistral, who shipped a decent Large 4 “Le Chonk” model on the new 3800 GB300 cluster funded by their recent Series D.” via Latent Space / AINews

    Reported bySimon Willison, Latent Space / AINews

    Discussed onHacker News 2,027 pts · 1,210 comments

    How 1 source put it
    mistral.ai First shared 2026-10-06
  10. Shared by 11

    Gemini 4 Argon: our next era of frontier intelligence

    Google announces Gemini 4 Argon, a frontier model launching at $2 per million input tokens and $10 per million output tokens. Details internal wins like a 40% quantum optimization beat, 1M token output limits, and a 91.7% LVBench score. cited

    Reported byLatent Space / AINews, The Rundown AI, Superhuman AI, TLDR AI, Ben's Bites, Don't Worry About the Vase, The Neuron, AI Breakfast, TheSequence, Turing Post

    Discussed onHacker News 1,702 pts · 1,190 comments

    Where it was shared
    blog.google announcement · 6 min read First shared 2026-09-30