Model Follower.

Up to date

Rebuilt
2026-11-15 07:14 UTC
Release radar
108 on watch
Hub snapshot
2026-11-15
Re-checked
daily · every number sourced and dated

Links / November 2026

Wednesday, 4 November 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 2026-11-04 23:59 UTC Status sealed

That day 36-hour window

  1. Shared by 5

    Siri AI is now open to all developers

    iOS 27.3 betas let developers integrate apps with Siri AI using App Intents and skills. Includes details on App Intents, skills for combining actions, and Spotlight semantic index access. cited

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

    Where it was shared
    developer.apple.com announcement · 3 min read First shared 2026-11-04
  2. Shared by 5

    Progress on provable inference for large language models

    Anthropic releases FlashProve, a system generating cryptographic proofs for a 400B-parameter model at 110 tokens/second. Includes executable code, Lean formalizations, and a protocol specification under the Apache 2.0 license. cited

    Reported byAlphaSignal, The Rundown AI, TLDR AI, AI Breakfast

    Discussed onHacker News 207 pts · 91 comments

    Where it was shared
    anthropic.com announcement · 20 min read First shared 2026-11-03
  3. Shared by 9

    Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights

    GPT-6 Astra edits 0.016% of Qwen3.5-0.8B weights to raise MATH-500 from 7.9% to 48.1% without training. Includes demos where Astra identifies games from raw weights and writes MinAtar agents beating PPO 100M steps. cited

    Reported byTuring Post, Latent Space / AINews, The Neuron, The Rundown AI, Superhuman AI, Simon Willison, AlphaSignal, AI Breakfast

    Discussed onHacker News 2,947 pts · 890 comments

    Where it was shared
    rikiyat.github.io paper · 18 min read First shared 2026-11-02
  4. Shared by 4

    The 'Astra can read and write weights to improve Qwen' post is misleading

    Matthew Honnibal argues the Astra Qwen post is misleading because it conflates raw weight reading with Python-based editing. Details a table showing only the Qwen experiment used Python to run the model, while true raw weight reading occurred on smaller networks. cited

    Reported bySimon Willison, TLDR AI, Latent Space / AINews

    Discussed onHacker News 1,029 pts · 207 comments

    Where it was shared
    honnibal.dev analysis · 4 min read First shared 2026-11-03
  5. Shared by 4

    OpenAI: ChatGPT Search Is About 20% of Google's, 250 Million People Speak With ChatGPT, and Chat Is Merging With Work

    “OpenAI said ChatGPT handles about one-fifth of Google's search-style queries, and weekly users rose from 1B to 1.2B in about 60 days.” via The Neuron

    Reported byThe Neuron, TLDR, TLDR AI, AI Breakfast

    How 1 source put it
    bigtechnology.com First shared 2026-11-04
  6. Shared by 3

    Introducing AI Discover: A New Place to Find Your Next Great Watch

    Netflix launches AI Discover, a new TV tab for a small group of U.S. members to browse AI-assisted themes. Includes a workflow where users pick themes like 'Something Light' or 'Family Movie Night' and tune picks via a 'Tune Your Recommendations' button. cited

    Reported byThe Neuron, The Rundown AI, Superhuman AI

    Where it was shared
    about.netflix.com announcement · 2 min read First shared 2026-11-04
  7. Shared by 6

    Anthropic warns that it's very close to RSI capability, aka self-building AI

    Anthropic's updated S-1 states it may imminently meet its definition of dramatically accelerated progress attributable to AI. Details the company's admission that it has already used AI to develop new models whose leaps in capability it didn't recognize at the time. cited

    Reported byForward Future, TLDR, TLDR AI, The Deep View, Mindstream

    Discussed onTechmeme

    Where it was shared
    techcrunch.com news report · 3 min read First shared 2026-11-02
  8. Shared by 2

    IST-DASLab squeezes a 763B DeepSeek coding model down to 72 GB at 1 bit per parameter

    “5. IST-DASLab squeezes a 763B DeepSeek coding model down to 72 GB at 1 bit per parameter” via AlphaSignal

    Reported byLatent Space / AINews, AlphaSignal

    AboutDeepSeek-V4.1-Flash-1017, DeepSeek-V4.1-Flash

    How 1 source put it
    huggingface.co First shared 2026-11-03
  9. Shared by 2

    Google later defended the move on X as “more choice, not less,” but its post only describes Search with AI Mode and AI…

    Reported byThe Rundown AI, Latent Space / AINews

    How 1 source put it
    x.com X post First shared 2026-11-04
  10. Shared by 3

    Google won’t give schools Search without AI

    “GOOGLE & AI IN SCHOOLS🏫 Google won’t give schools Search without AIImage source: GoogleThe Rundown: Google announced that K-12 student accounts will start with Search, Gemini, and Gemini Notebook all off early next year, potentially forcing districts that want AI out of classrooms to also give up Google Search for their students.The details: Education VP…” via The Rundown AI

    Reported byThe Rundown AI, TLDR

    Discussed onHacker News 665 pts · 460 comments

    How 1 source put it
    • The Rundown AI2026-11-04issue ↗ GOOGLE & AI IN SCHOOLS🏫 Google won’t give schools Search without AIImage source: GoogleThe Rundown: Google announced that K-12 student accounts will start with Search, Gemini, and Gemini Notebook all off early next year, potentially forcing districts that want AI out of classrooms to also give up Google Search for their students.The details: Education VP…
    • TLDR2026-11-04issue ↗
    nytimes.com First shared 2026-11-03
  11. Shared by 2

    Kimi K3.1 Pro model weights are out

    “Moonshot AI released full open weights for Kimi K3.1 Pro on Hugging Face” via AlphaSignal

    Reported byAlphaSignal

    Discussed onHacker News 479 pts · 135 comments

    huggingface.co First shared 2026-11-04
  12. Shared by 4

    OpenAI is finally merging ChatGPT's Chat and Work tabs

    OpenAI merges Chat and Work tabs in the ChatGPT desktop app for select users. Details how files and plugins stay together while Codex remains separate and shares the same token pool. cited

    Reported byForward Future, The Deep View, TLDR AI, Superhuman AI

    Where it was shared
    techcrunch.com news report · 1 min read First shared 2026-11-03
  13. Shared by 2

    Latent Sparse Attention: Towards Subquadratic Ultra-Long Context Attention

    Latent Sparse Attention clusters keys hierarchically to enable subquadratic attention at ultra-long contexts above 1M tokens. Includes results showing Distilled-LSA speeds up one attention layer by 22.7x in prefill and 4.6x in decode on GLM-5.3-FLASH at 10M tokens. cited

    Reported byAlphaSignal, Latent Space / AINews

    Where it was shared
    arxiv.org paper · 2 min read First shared 2026-11-03
  14. Shared by 2

    Designing responsible AI for learning

    Google turns off Gemini and Search by default for K-12 accounts until administrators enable them. Includes a pilot in select U.S. schools later this month and global availability of Guided Learning in Classroom today. cited

    Reported byThe Neuron, The Rundown AI

    Where it was shared
    blog.google announcement · 2 min read First shared 2026-11-04
  15. Shared by 2

    Plugin auto-connect: Plugins now connect automatically via MCP servers that trust Claude as an issuer (ClaudeDevs).

    Reported byAI Breakfast, Latent Space / AINews

    How 1 source put it
    x.com X post First shared 2026-11-04
  16. Shared by 2

    How much do tokens really cost?

    “How much do tokens really cost? (7 minute read) It appears that the cost of running servers is significantly cheaper than what providers are charging, assuming high capacity.” via TLDR

    Reported byTLDR, TLDR AI

    How 1 source put it
    • TLDR2026-11-04issue ↗ How much do tokens really cost? (7 minute read) It appears that the cost of running servers is significantly cheaper than what providers are charging, assuming high capacity.
    • TLDR AI2026-11-04issue ↗
    substack.hadley.nz 5 min read First shared 2026-11-04
  17. Shared by 5

    OpenAI grades its AI on everyday tasks

    “OpenAI unveiled LifeVal, a benchmark grading models on everyday personal-assistant tasks.” via AlphaSignal

    Reported byAlphaSignal, The Rundown AI, The Deep View, TLDR AI, AI Breakfast

    How 1 source put it
    openai.com First shared 2026-11-02
  18. Shared by 2

    Learnings from Auditing Public Evaluations

    “Learnings from Auditing Public Evaluations (8 minute read) The article outlines issues faced during public evaluations of LLMs and details improvements made to benchmarks such as AUTOMATIONBENCH and OFFICEQA PRO V2. They replaced ambiguous prose descriptions with JSON schemas for better performance evaluations and introduced new grading systems to improve…” via TLDR AI

    Reported byAlphaSignal, TLDR AI

    How 1 source put it
    • AlphaSignal2026-11-03issue ↗
    • TLDR AI2026-11-04issue ↗ Learnings from Auditing Public Evaluations (8 minute read) The article outlines issues faced during public evaluations of LLMs and details improvements made to benchmarks such as AUTOMATIONBENCH and OFFICEQA PRO V2. They replaced ambiguous prose descriptions with JSON schemas for better performance evaluations and introduced new grading systems to improve…
    primeintellect.ai 7 min read First shared 2026-11-03
  19. Shared by 1

    As a Summit it is meant to be more intimate: single track, invite-only for the top 1000 agentic engineers in the world…

    “…is the last flagship AIE of the year (there are also 2 partner conferences coming up - AIEi Shanghai takes place next week, and AIEi Sydney is a side event of NeurIPS in December). As a Summit it is meant to be more intimate: single track, invite-only for the top 1000 agentic engineers in the world (over 100k tuned into our Software Factories track at…” via Latent Space / AINews

    Reported byLatent Space / AINews

    ai.engineer First shared 2026-10-30
  20. Shared by 2

    OpenAI quadruples bug bounty payouts, now offering up to $172,000 per critical find

    “The company has also expanded its security leadership. OpenAI hired former Palantir chief information security officer Dane Stuckey to work alongside security chief Matt Knight. Its sprint announcement describes a repeated cycle of authorized testing, remediation, and retesting.” via AlphaSignal

    Reported byAlphaSignal, The Deep View

    How 1 source put it
    • AlphaSignal2026-11-03issue ↗ The company has also expanded its security leadership. OpenAI hired former Palantir chief information security officer Dane Stuckey to work alongside security chief Matt Knight. Its sprint announcement describes a repeated cycle of authorized testing, remediation, and retesting.
    • The Deep View2026-11-04issue ↗
    openai.com First shared 2026-11-03

The week to that day two or more sources

  1. Shared by 9

    Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights

    GPT-6 Astra edits 0.016% of Qwen3.5-0.8B weights to raise MATH-500 from 7.9% to 48.1% without training. Includes demos where Astra identifies games from raw weights and writes MinAtar agents beating PPO 100M steps. cited

    Reported byTuring Post, Latent Space / AINews, The Neuron, The Rundown AI, Superhuman AI, Simon Willison, AlphaSignal, AI Breakfast

    Discussed onHacker News 2,947 pts · 890 comments

    Where it was shared
    rikiyat.github.io paper · 18 min read First shared 2026-11-02
  2. Shared by 5

    Siri AI is now open to all developers

    iOS 27.3 betas let developers integrate apps with Siri AI using App Intents and skills. Includes details on App Intents, skills for combining actions, and Spotlight semantic index access. cited

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

    Where it was shared
    developer.apple.com announcement · 3 min read First shared 2026-11-04
  3. Shared by 5

    Progress on provable inference for large language models

    Anthropic releases FlashProve, a system generating cryptographic proofs for a 400B-parameter model at 110 tokens/second. Includes executable code, Lean formalizations, and a protocol specification under the Apache 2.0 license. cited

    Reported byAlphaSignal, The Rundown AI, TLDR AI, AI Breakfast

    Discussed onHacker News 207 pts · 91 comments

    Where it was shared
    anthropic.com announcement · 20 min read First shared 2026-11-03
  4. Shared by 4

    The 'Astra can read and write weights to improve Qwen' post is misleading

    Matthew Honnibal argues the Astra Qwen post is misleading because it conflates raw weight reading with Python-based editing. Details a table showing only the Qwen experiment used Python to run the model, while true raw weight reading occurred on smaller networks. cited

    Reported bySimon Willison, TLDR AI, Latent Space / AINews

    Discussed onHacker News 1,029 pts · 207 comments

    Where it was shared
    honnibal.dev analysis · 4 min read First shared 2026-11-03
  5. Shared by 6

    Anthropic warns that it's very close to RSI capability, aka self-building AI

    Anthropic's updated S-1 states it may imminently meet its definition of dramatically accelerated progress attributable to AI. Details the company's admission that it has already used AI to develop new models whose leaps in capability it didn't recognize at the time. cited

    Reported byForward Future, TLDR, TLDR AI, The Deep View, Mindstream

    Discussed onTechmeme

    Where it was shared
    techcrunch.com news report · 3 min read First shared 2026-11-02
  6. Shared by 15

    Gemini 4 Carbon: our next frontier model

    “Google announced Gemini 4 Carbon, its new frontier model focused on cybersecurity and long-horizon agentic coding.” via AlphaSignal

    Reported byAlphaSignal, Latent Space / AINews, The Neuron, The Rundown AI, Forward Future, Superhuman AI, Ben's Bites, TLDR AI, AI Breakfast, NLP Newsletter, What's Up in AI, TheSequence, FryAI, Turing Post

    Discussed onHacker News 1,019 pts · 803 comments

    How 1 source put it
    blog.google 4 min read First shared 2026-10-28
  7. Shared by 5

    OpenAI grades its AI on everyday tasks

    “OpenAI unveiled LifeVal, a benchmark grading models on everyday personal-assistant tasks.” via AlphaSignal

    Reported byAlphaSignal, The Rundown AI, The Deep View, TLDR AI, AI Breakfast

    How 1 source put it
    openai.com First shared 2026-11-02
  8. Shared by 6

    We’ve received our first response from the Project Suncatcher prototype satellite.

    Google received the first AI response from its Project Suncatcher satellite after uplinking a prompt to the onboard Gemma 4 model. Includes the haiku generated by the model as it passed Svalbard and plans to test thermal stress and radiation exposure next. cited

    Reported byThe Rundown AI, The Deep View, TLDR AI, AI Breakfast, FryAI, AlphaSignal

    Aboutgemma-4-26B-A4B-it-assistant, gemma-4-31B-it-assistant

    Where it was shared
    blog.google news report · 1 min read First shared 2026-11-02
  9. Shared by 4

    OpenAI: ChatGPT Search Is About 20% of Google's, 250 Million People Speak With ChatGPT, and Chat Is Merging With Work

    “OpenAI said ChatGPT handles about one-fifth of Google's search-style queries, and weekly users rose from 1B to 1.2B in about 60 days.” via The Neuron

    Reported byThe Neuron, TLDR, TLDR AI, AI Breakfast

    How 1 source put it
    bigtechnology.com First shared 2026-11-04
  10. Shared by 3

    Introducing AI Discover: A New Place to Find Your Next Great Watch

    Netflix launches AI Discover, a new TV tab for a small group of U.S. members to browse AI-assisted themes. Includes a workflow where users pick themes like 'Something Light' or 'Family Movie Night' and tune picks via a 'Tune Your Recommendations' button. cited

    Reported byThe Neuron, The Rundown AI, Superhuman AI

    Where it was shared
    about.netflix.com announcement · 2 min read First shared 2026-11-04