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

Sunday, 18 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 2026-10-18 23:59 UTC Status sealed Rebuilt from issues on file

Reconstructed after the day from the issues on file (dates as published). An issue counts on the date it carries, even if it was collected later.

That day 36-hour window

  1. Shared by 7

    Introducing Muse Spark Open, The Most Capable US Open‑Weight Model to Date

    Meta releases Muse Spark Open, an 816B-parameter multimodal model designed for coding and agentic work. Details how two Apple M5 Ultra Mac Studios run the model locally via MLX and lists hosted inference partners. cited

    Reported byThe Rundown AI, The Neuron, Superhuman AI, TLDR AI, AI Breakfast, Interconnects, TheSequence

    Where it was shared
    research.meta.ai announcement · 3 min read First shared 2026-10-16
  2. Shared by 1

    Release: llm-tools-datasette 0.1.1 When I switched LLM over to httpx2 in LLM 0.33 I forget that this would break any…

    Reported bySimon Willison

    github.com First shared 2026-10-18
  3. Shared by 1

    When I switched LLM over to httpx2 in LLM 0.33 I forget that this would break any plugins which used httpx without…

    Reported bySimon Willison

    github.com First shared 2026-10-18
  4. Shared by 1

    Changelog#

    “When I switched LLM over to httpx2 in LLM 0.33 I forget that this would break any plugins which used httpx without directly depending on it. llm-tools-datasette was one of those plugins, now fixed.” via Simon Willison

    Reported bySimon Willison

    llm.datasette.io 52 min read First shared 2026-10-18
  5. Shared by 1

    Release: llm-echo 0.5 I'm working towards a new LLM release, and testing PRs caused me to revisit the llm-echo debug…

    Reported bySimon Willison

    github.com First shared 2026-10-18
  6. Shared by 9

    Muse Wonder: A Frontier Model For Personal & Professional AI Agents

    Meta releases Muse Wonder and Muse Wonder Full, its most capable frontier models for personal and professional agents. Details Active Memory and Active Focus features, plus benchmark results comparing performance on law, finance, and coding tasks. cited

    Reported byAI Breakfast, Simon Willison, The Rundown AI, Latent Space / AINews, The Neuron, TLDR AI, Ben's Bites, Interconnects, TheSequence

    Where it was shared
    research.meta.ai announcement · 4 min read First shared 2026-10-12
  7. Shared by 1

    Hybrid Reasoning Agents: LLM agents that think in latent space, act through text

    Discussed onHacker News 520 pts · 239 comments

    arxiv.org First shared 2026-10-18
  8. Shared by 1

    Rivian’s own assistant already shows the basic idea: connect your calendar, then ask it to move a meeting or navigate…

    “Think of the direction as an ecosystem of assistants in your car, with software from different companies that can handle tasks for you. Rivian’s own assistant already shows the basic idea: connect your calendar, then ask it to move a meeting or navigate to your next appointment. The new part is opening the car to outside agents, too.” via The Neuron

    Reported byThe Neuron

    rivian.com First shared 2026-10-18
  9. Shared by 1

    If you don't run your own LLM proxy, how do you know whats going on?

    Discussed onHacker News 302 pts · 219 comments

    jrdsgl.com First shared 2026-10-18
  10. Shared by 2

    Claude Opus 5 scored 100% on month-end close tasks where 12 licensed CPAs averaged about 37%

    “Claude Opus 5 scored 100% on month-end close tasks where 12 licensed CPAs averaged about 37% (Mercor)” via AI Breakfast

    Reported byAI Breakfast, The Neuron

    How 1 source put it
    mercor.com First shared 2026-10-05
  11. Shared by 1

    Click the header image to go straight to YouTube!Welcome, humans.First up: we go LIVE in 5 minutes with Alex Lavaee…

    Reported byThe Neuron

    web.cvent.com First shared 2026-10-08
  12. Shared by 2

    Analyzing technical documents with Deep Review

    “Google's Deep Review is an agent AI system engineer designed to provide rigorous feedback and technical verification on complex documents. It was initially designed for computer science academic paper analysis, but its capabilities have since expanded to reviewing documents across many domains.…” via TLDR AI

    Reported byTLDR AI, The Neuron

    How 1 source put it
    • TLDR AI2026-10-15issue ↗ Google's Deep Review is an agent AI system engineer designed to provide rigorous feedback and technical verification on complex documents. It was initially designed for computer science academic paper analysis, but its capabilities have since expanded to reviewing documents across many domains.…
    • The Neuron2026-10-16issue ↗
    research.google First shared 2026-10-15
  13. Shared by 1

    Legal experts told Nieman Lab

    “Here’s what changed: Meta’s business terms take effect October 30 and authorize scraping websites associated with businesses using its products. That means automatically collecting content from those sites to develop and train Meta products. Legal experts told Nieman Lab that could include training its AI on publishers’ articles.” via The Neuron

    Reported byThe Neuron

    niemanlab.org First shared 2026-10-18
  14. Shared by 1

    Five major publishers sued Meta in May, alleging it pirated millions of works to train its Llama AI models.

    “… gets permission to collect content, using it to train AI is another legal question: does it have the right to learn from that work? Five major publishers sued Meta in May, alleging it pirated millions of works to train its Llama AI models. That’s a separate case, but it shows why a new set of terms doesn’t end the argument over whose work AI can learn…” via The Neuron

    Reported byThe Neuron

    theguardian.com First shared 2026-10-18
  15. Shared by 1

    Google is paying publishers.

    “On October 15, it said more than 400 publications participate in its News AI pilot. CNN, Axios, and the Financial Times are among the new partners. Their content powers things like article summaries and audio briefings.” via The Neuron

    Reported byThe Neuron

    blog.google First shared 2026-10-18
  16. Shared by 1

    Cloudflare is testing payment for AI use.

    “Its Pay Per Use beta, announced September 30, lets publishers accept offers from AI companies. The buyer sets the price and what it wants to use the content for, then reports its own usage. Cloudflare handles the bills and payouts.” via The Neuron

    Reported byThe Neuron

    blog.cloudflare.com First shared 2026-10-18
  17. Shared by 1

    ChatGPT's Chrome extension can help you compare information across open tabs without constantly switching between them.

    Reported byThe Neuron

    chromewebstore.google.com First shared 2026-10-18
  18. Shared by 1

    Why Successful AI Pilots Still Struggle to Reach Enterprise Scale. Read more.

    Reported byThe Neuron

    eweek.com First shared 2026-10-18
  19. Shared by 1

    helps you find specific moments in videos using plain-English searches.

    Reported byThe Neuron

    playground.twelvelabs.io First shared 2026-10-18
  20. Shared by 1

    proposed new export restrictions covering 23 categories of AI technology, with an exemption for technology already…

    Reported byThe Neuron

    wsj.com First shared 2026-10-18

The week to that day two or more sources

  1. Shared by 7

    Introducing Muse Spark Open, The Most Capable US Open‑Weight Model to Date

    Meta releases Muse Spark Open, an 816B-parameter multimodal model designed for coding and agentic work. Details how two Apple M5 Ultra Mac Studios run the model locally via MLX and lists hosted inference partners. cited

    Reported byThe Rundown AI, The Neuron, Superhuman AI, TLDR AI, AI Breakfast, Interconnects, TheSequence

    Where it was shared
    research.meta.ai announcement · 3 min read First shared 2026-10-16
  2. Shared by 9

    Muse Wonder: A Frontier Model For Personal & Professional AI Agents

    Meta releases Muse Wonder and Muse Wonder Full, its most capable frontier models for personal and professional agents. Details Active Memory and Active Focus features, plus benchmark results comparing performance on law, finance, and coding tasks. cited

    Reported byAI Breakfast, Simon Willison, The Rundown AI, Latent Space / AINews, The Neuron, TLDR AI, Ben's Bites, Interconnects, TheSequence

    Where it was shared
    research.meta.ai announcement · 4 min read First shared 2026-10-12
  3. Shared by 9

    Redesigning source control at Anthropic

    Anthropic replaced git with Scion, merging over 30,000 code changes daily after migrating their monorepo in five months. Details how Scion uses always-online agents, server-side commits, and sparse checkouts to avoid storing full repo copies locally. cited

    Reported byAlphaSignal, Simon Willison, Superhuman AI, AI Breakfast, The Neuron, Latent Space / AINews, Don't Worry About the Vase, Ben's Bites

    Discussed onHacker News 898 pts · 1,013 comments

    Where it was shared
    anthropic.com announcement · 16 min read First shared 2026-10-13
  4. Shared by 5

    An Update on Character.ai

    Character.ai initiated a voluntary Chapter 11 bankruptcy process to facilitate a court-supervised sale of the company. Details the timeline expecting completion by the end of November or perhaps early December 2026 and confirms subscriptions remain active. cited

    Reported byThe Rundown AI, TLDR AI, Data Points, TheSequence

    Discussed onHacker News 336 pts · 296 comments

    Where it was shared
    blog.character.ai news report · 2 min read First shared 2026-10-13
  5. Shared by 6

    Gemini 4 Argon: Built to solve your hardest problems

    Gemini 4 Argon launches globally today for Pro and Ultra subscribers via the Gemini app and API. Includes internal tests where Argon watched a human play cornhole, built simulation code, and sank seven consecutive shots on a robot. cited

    Reported byAlphaSignal, The Rundown AI, The Neuron, Superhuman AI, Forward Future, Ben's Bites

    Where it was shared
    blog.google announcement · 10 min read First shared 2026-10-13
  6. Shared by 12

    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, Last Week in AI, TheSequence, Import AI, Turing Post, The Batch, Interconnects

    Discussed onHacker News 522 pts · 5 comments

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

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

    “Anthropic released new research demonstrating that small AI models are getting closer to being able to copy themselves between everyday gadgets like smart ovens or TVs.” via The Rundown AI

    Reported byThe Rundown AI, The Neuron, TLDR AI, Latent Space / AINews, AI Breakfast, Don't Worry About the Vase

    Discussed onHacker News 118 pts · 116 comments

    How 1 source put it
    anthropic.com 22 min read First shared 2026-10-12
  8. Shared by 3

    Inside Microsoft's Frontier Pre-Training Stack

    “Microsoft detailed the training of MAI-Base-2, a 2.8T-parameter mixture-of-experts model with 80B active parameters that it said surpassed every open-weight model in its held-out pre-training evaluations. The deep dive covers its data strategy, compute efficiency, full-stack infrastructure, and…” via TLDR AI

    Reported byTLDR AI, AI Breakfast, Interconnects

    How 1 source put it
    • TLDR AI2026-10-16issue ↗ Microsoft detailed the training of MAI-Base-2, a 2.8T-parameter mixture-of-experts model with 80B active parameters that it said surpassed every open-weight model in its held-out pre-training evaluations. The deep dive covers its data strategy, compute efficiency, full-stack infrastructure, and…
    • AI Breakfast2026-10-16issue ↗
    • Interconnects2026-10-17issue ↗
    microsoft.ai First shared 2026-10-16
  9. Shared by 3

    Jev: For the builders AND dreamers

    TypeSafe releases Jev 1.14.0 with experimental image support, bounding box outputs, and longer context behind API flags. Details how to enable multimodality and extended context via specific header flags, noting models perform worse on these tasks than pure text. cited

    Reported byTLDR AI, Ben's Bites, TheSequence

    Where it was shared
    typesafe.ai announcement · 4 min read First shared 2026-10-15
  10. Shared by 3

    Anthropic update on independent evaluators

    Anthropic embedded six METR employees with hardware and access matching its internal risk teams. Details how Anthropic can require redactions for confidential IP while METR publishes findings without approval. cited

    Reported byThe Rundown AI, The Neuron, TLDR AI

    Where it was shared
    anthropic.com announcement · 4 min read First shared 2026-10-16