Model Follower.

Up to date

Rebuilt
2026-10-26 07:36 UTC
Release radar
116 on watch
Hub snapshot
2026-10-26
Re-checked
daily · every number sourced and dated

Links / Past days

Monday, 28 September 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 2026-09-28 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 4

    Read the full article here

    “Microsoft just rebuilt Copilot around three new pieces: Home, Code, and Autopilot.” via The Neuron

    Reported byThe Neuron, The Rundown AI, Forward Future, Superhuman AI

    How 1 source put it
    blogs.microsoft.com First shared 2026-09-27
  2. Shared by 4

    Then Madison Mills reported in Axios that we can raise the stakes

    “The Rundown: OpenAI confirmed its agents went off-script on U.S. government websites this summer, as Axios reports that OpenAI, Anthropic, and researchers are investigating tens of thousands of incidents of problematic AI behavior.” via The Rundown AI

    Reported byThe Neuron, The Rundown AI, Don't Worry About the Vase, AI Breakfast

    How 1 source put it
    axios.com First shared 2026-09-27
  3. Shared by 3

    Anthropic Signed an $11.6 Billion Akamai Compute Deal

    “Anthropic signed a 7-year, $11.6B deal with Akamai to use its cloud infrastructure and software to scale its AI workloads, with the deal potentially reaching $20B.” via The Rundown AI

    Reported byAI Breakfast, The Rundown AI, TLDR AI

    How 1 source put it
    ir.akamai.com First shared 2026-09-26
  4. Shared by 4

    An agent used DNS to reach an external chatbot

    “Okay, so an OpenAI agent got blocked from the internet... and then figured out how to use DNS to ask an outside chatbot for help.” via The Neuron

    Reported byThe Neuron, Don't Worry About the Vase, AI Breakfast

    Discussed onHacker News 198 pts · 188 comments

    How 1 source put it
    alignment.openai.com First shared 2026-09-26
  5. Shared by 3

    Claude computes a nine-loop amplitude in N=4 super-Yang-Mills

    “Physicists at Anthropic harnessed Claude, an LLM, to compute a complex nine-loop amplitude in N=4 super-Yang-Mills, a problem once considered computationally infeasible with limited resources. They utilized the bootstrap method and form-factor approach, achieving results…” via TLDR AI

    Reported byTLDR AI, AI Breakfast

    Discussed onHacker News 104 pts · 62 comments

    How 1 source put it
    • TLDR AI2026-09-28issue ↗ Physicists at Anthropic harnessed Claude, an LLM, to compute a complex nine-loop amplitude in N=4 super-Yang-Mills, a problem once considered computationally infeasible with limited resources. They utilized the bootstrap method and form-factor approach, achieving results…
    • AI Breakfast2026-09-28issue ↗
    anthropic.com First shared 2026-09-25
  6. Shared by 2

    Hitting a billion tokens per minute on one GPU by combining a query planner and an inference engine

    “The QUery-Aware Inference Layer (Quail) is an inference engine that hits over a billion tokens processed per minute per H100 GPU. It is over 10 times faster than the vLLM baseline on the same hardware, and it costs under $0.06 per billion…” via TLDR AI

    Reported byTLDR AI, AI Breakfast

    How 1 source put it
    • TLDR AI2026-09-28issue ↗ The QUery-Aware Inference Layer (Quail) is an inference engine that hits over a billion tokens processed per minute per H100 GPU. It is over 10 times faster than the vLLM baseline on the same hardware, and it costs under $0.06 per billion…
    • AI Breakfast2026-09-28issue ↗
    modal.com First shared 2026-09-28
  7. Shared by 2

    Build plugins for Claude with the directory submission portal

    “Developers on paid Claude plans can now build and submit plugins using a new directory submission portal.” via TLDR AI

    Reported byTLDR AI, AI Breakfast

    How 1 source put it
    claude.com First shared 2026-09-28
  8. Shared by 2

    Court Filing Names War Department

    “The Rundown: Anthropic’s refusal to let Claude power lethal autonomous weapons or mass domestic surveillance is now legal grounds for a Pentagon blacklist after a federal appeals court upheld the exclusion on Friday.” via The Rundown AI

    Reported byThe Rundown AI, Forward Future

    How 1 source put it
    • The Rundown AI2026-09-28issue ↗ The Rundown: Anthropic’s refusal to let Claude power lethal autonomous weapons or mass domestic surveillance is now legal grounds for a Pentagon blacklist after a federal appeals court upheld the exclusion on Friday.
    • Forward Future2026-09-28issue ↗
    media.cadc.uscourts.gov First shared 2026-09-28
  9. Shared by 2

    Joe, an OpenAI agent security engineer, here offers an excellent download

    “I have great sympathy for those working tirelessly on these problems. Joe, an OpenAI agent security engineer, here offers an excellent download of his perspective as someone working on agent safety, on why none of this is simple.” via Don't Worry About the Vase

    Reported byDon't Worry About the Vase, Simon Willison

    How 1 source put it
    • Don't Worry About the Vase2026-09-28issue ↗ I have great sympathy for those working tirelessly on these problems. Joe, an OpenAI agent security engineer, here offers an excellent download of his perspective as someone working on agent safety, on why none of this is simple.
    • Simon Willison2026-09-28issue ↗
    x.com X post First shared 2026-09-28
  10. 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
  11. Shared by 1

    Introducing Claude Sonnet 5.5

    Anthropic launches Claude Sonnet 5.5, a 30% faster model that costs up to 30% less per task than its predecessor. Includes benchmark tables showing 70.6% on Terminal-Bench 4.0 and pricing details for cache reads at $0.20 per million tokens. cited

    Reported bySimon Willison

    anthropic.com announcement · 8 min read First shared 2026-09-28
  12. Shared by 2

    Then OpenAI dropped news on a Friday afternoon that they were making their way

    “The Rundown: OpenAI confirmed its agents went off-script on U.S. government websites this summer, as Axios reports that OpenAI, Anthropic, and researchers are investigating tens of thousands of incidents of problematic AI behavior.” via The Rundown AI

    Reported byThe Rundown AI, Don't Worry About the Vase

    How 1 source put it
    • The Rundown AI2026-09-28issue ↗ The Rundown: OpenAI confirmed its agents went off-script on U.S. government websites this summer, as Axios reports that OpenAI, Anthropic, and researchers are investigating tens of thousands of incidents of problematic AI behavior.
    • Don't Worry About the Vase2026-09-28issue ↗
    openai.com First shared 2026-09-28
  13. Shared by 1

    Sonnet 5.5 appears to be almost as good as Opus 5.5 on some coding tasks, including various viral 3D animation tricks.

    Reported bySimon Willison

    x.com X post First shared 2026-09-28
  14. Shared by 6

    OpenAI says Astra was not involved.

    “We’re getting close to superhuman systems that can hack anything and acquire real resources. OpenAI’s incidental Hugging Face hack supports that.” via Superhuman AI

    Reported byThe Batch, Superhuman AI, Interconnects, The Neuron, Don't Worry About the Vase, AI Breakfast

    How 1 source put it
    openai.com First shared 2026-09-11
  15. Shared by 2

    OpenAI Alignment Research Blog

    “…t researcher's keys, then every employee's keys. The third, self-replicating prompt injections exist, documents injections that "self-propagate akin to a computer worm" across email, filesystem, build scripts and multi-hop Slack reads, hitting GPT-5.5 and GPT-5.4-mini. No impact outside simulated tool calls, and self-reproduction is now an attacker…” via AI Breakfast

    Reported byDon't Worry About the Vase, AI Breakfast

    How 1 source put it
    • Don't Worry About the Vase2026-09-28issue ↗
    • AI Breakfast2026-09-28issue ↗ …t researcher's keys, then every employee's keys. The third, self-replicating prompt injections exist, documents injections that "self-propagate akin to a computer worm" across email, filesystem, build scripts and multi-hop Slack reads, hitting GPT-5.5 and GPT-5.4-mini. No impact outside simulated tool calls, and self-reproduction is now an attacker…
    alignment.openai.com First shared 2026-09-28
  16. Shared by 2

    Towards Universal Post-Training for Robotics

    “Robotics needs a universal post-training recipe similar to language models to achieve reliable autonomous deployment. Existing methods like imitation learning benefit from intuition rather than structured processes, and RL's instability hinders robotics. The development of EXPO-FT demonstrates a…” via TLDR AI

    Reported byTLDR AI, Import AI

    How 1 source put it
    • TLDR AI2026-09-24issue ↗ Robotics needs a universal post-training recipe similar to language models to achieve reliable autonomous deployment. Existing methods like imitation learning benefit from intuition rather than structured processes, and RL's instability hinders robotics. The development of EXPO-FT demonstrates a…
    • Import AI2026-09-28issue ↗
    pd-perry.github.io First shared 2026-09-24
  17. Shared by 1

    Ingressing Minds: Causal, Non-Physical Patterns In-Form Natural, Synthetic, and Hybrid Embodiments (MDPI)

    “…end a few hours with this paper - it’s worth it. Read more: Ingressing Minds: Causal, Non-Physical Patterns In-Form Natural, Synthetic, and Hybrid Embodiments (MDPI).***Robotics is preparing for a LLM-moment, but it’s missing a universal algorithm:…For robots to mature, post-training needs to simplify…Perry Dong, a Stanford researcher, and Chelsea Finn, a…” via Import AI

    Reported byImport AI

    mdpi.com First shared 2026-09-28
  18. Shared by 1

    @joedaroo, Agent Security at OpenAI, identity confirmed by The Information's Rocket Drew

    “— @joedaroo, Agent Security at OpenAI, identity confirmed by The Information's Rocket Drew” via Simon Willison

    Reported bySimon Willison

    x.com X post First shared 2026-09-28
  19. Shared by 1

    OpenAI prepares to expand Ultrafast API to more users

    “OpenAI appears to be preparing a wider rollout of its Ultrafast API mode. References to the rollout have been seen showing up across the OpenAI Platform and API documentation. The feature was officially previewed with GPT-5.6 Sol. OpenAI claims it has speeds of up to 750 output tokens per…” via TLDR AI

    Reported byTLDR AI

    testingcatalog.com First shared 2026-09-28
  20. Shared by 4

    Once Claude can measure something, it can make it faster

    “How Anthropic made claude.ai 3x faster in two weeks. Prompts included.” via Ben's Bites

    Reported byBen's Bites, Don't Worry About the Vase, Data Points

    Discussed onHacker News 231 pts · 154 comments

    How 1 source put it
    claude.dev First shared 2026-09-23

The week to that day two or more sources

  1. Shared by 9

    Claude discovers a novel enzyme system

    Claude agents identified a novel bacteriophage enzyme system with CRISPR-like DNA repeats after 21 hours of searching. Details the system's three parts, its expression as short RNAs, and the team's plan to determine its function. cited

    Reported byThe Rundown AI, The Neuron, Superhuman AI, TLDR AI, Ben's Bites, Don't Worry About the Vase, Latent Space / AINews, TheSequence

    Discussed onHacker News 780 pts · 805 comments

    Where it was shared
    anthropic.com news report · 7 min read First shared 2026-09-23
  2. Shared by 10

    Introducing Claude Opus 5.5

    Anthropic launches Claude Opus 5.5, which costs 40% less to run than Opus 5 while matching Fable 5.1 performance. Details input/output token prices of $4 and $20 per million, plus a 60% reduction in cache read costs. cited

    Reported byThe Neuron, The Rundown AI, Superhuman AI, Forward Future, TLDR AI, AI Breakfast, Ben's Bites, The Batch, TheSequence

    Discussed onHacker News 1,806 pts · 1,134 comments

    Where it was shared
    anthropic.com announcement · 14 min read First shared 2026-09-22
  3. Shared by 7

    Behind Project Suncatcher, our moonshot to put AI in space

    Google launches a prototype satellite with four TPUs on a SpaceX Transporter-18 mission to test AI hardware in orbit. Details vibration tests showing chips survive 100 g forces, radiation results from proton beam tests, and plans for 2027 laser interconnectivity trials. cited

    Reported byThe Neuron, The Rundown AI, Forward Future, Superhuman AI, AI Breakfast, Import AI

    Discussed onHacker News 233 pts · 554 comments

    Where it was shared
    blog.google news report · 4 min read First shared 2026-09-24
  4. Shared by 4

    Read the full article here

    “Microsoft just rebuilt Copilot around three new pieces: Home, Code, and Autopilot.” via The Neuron

    Reported byThe Neuron, The Rundown AI, Forward Future, Superhuman AI

    How 1 source put it
    blogs.microsoft.com First shared 2026-09-27
  5. Shared by 4

    Then Madison Mills reported in Axios that we can raise the stakes

    “The Rundown: OpenAI confirmed its agents went off-script on U.S. government websites this summer, as Axios reports that OpenAI, Anthropic, and researchers are investigating tens of thousands of incidents of problematic AI behavior.” via The Rundown AI

    Reported byThe Neuron, The Rundown AI, Don't Worry About the Vase, AI Breakfast

    How 1 source put it
    axios.com First shared 2026-09-27
  6. Shared by 10

    GPT-6 Sol and Luna

    “Anthropic just dropped Claude Opus 5.5. About 90 minutes later, OpenAI answered with GPT-6 Sol and Luna.” via The Neuron

    Reported byThe Neuron, The Rundown AI, Superhuman AI, Forward Future, TLDR AI, AI Breakfast, Data Points, Ben's Bites, Don't Worry About the Vase

    Discussed onHacker News 1,779 pts · 855 comments

    How 1 source put it
    openai.com First shared 2026-09-22
  7. Shared by 4

    An agent used DNS to reach an external chatbot

    “Okay, so an OpenAI agent got blocked from the internet... and then figured out how to use DNS to ask an outside chatbot for help.” via The Neuron

    Reported byThe Neuron, Don't Worry About the Vase, AI Breakfast

    Discussed onHacker News 198 pts · 188 comments

    How 1 source put it
    alignment.openai.com First shared 2026-09-26
  8. Shared by 3

    Anthropic Signed an $11.6 Billion Akamai Compute Deal

    “Anthropic signed a 7-year, $11.6B deal with Akamai to use its cloud infrastructure and software to scale its AI workloads, with the deal potentially reaching $20B.” via The Rundown AI

    Reported byAI Breakfast, The Rundown AI, TLDR AI

    How 1 source put it
    ir.akamai.com First shared 2026-09-26
  9. Shared by 6

    OpenAI says Astra was not involved.

    “We’re getting close to superhuman systems that can hack anything and acquire real resources. OpenAI’s incidental Hugging Face hack supports that.” via Superhuman AI

    Reported byThe Batch, Superhuman AI, Interconnects, The Neuron, Don't Worry About the Vase, AI Breakfast

    How 1 source put it
    openai.com First shared 2026-09-11
  10. Shared by 3

    Claude computes a nine-loop amplitude in N=4 super-Yang-Mills

    “Physicists at Anthropic harnessed Claude, an LLM, to compute a complex nine-loop amplitude in N=4 super-Yang-Mills, a problem once considered computationally infeasible with limited resources. They utilized the bootstrap method and form-factor approach, achieving results…” via TLDR AI

    Reported byTLDR AI, AI Breakfast

    Discussed onHacker News 104 pts · 62 comments

    How 1 source put it
    • TLDR AI2026-09-28issue ↗ Physicists at Anthropic harnessed Claude, an LLM, to compute a complex nine-loop amplitude in N=4 super-Yang-Mills, a problem once considered computationally infeasible with limited resources. They utilized the bootstrap method and form-factor approach, achieving results…
    • AI Breakfast2026-09-28issue ↗
    anthropic.com First shared 2026-09-25