Post-Cutoff

Posts that mattered

1,633 posts by 887 authors: the X posts, essays and blog posts our events cite, newest first.

October 2026, continued

  1. The Transcript@TheTranscript_X

    The Transcript: Zuckerberg says ‘we were not happy with the trajectory that we were on with Llama 4’

    Transcribed Zuckerberg quote from the Cheeky Pint interview on why Meta rebuilt its model team after Llama 4.

  2. Torsten Hoefler 🇨🇭@thoeflerX

    Torsten Hoefler on Maia 300: ‘Our SDLA is a great architecture for LLMs’

    First-hand comment from an ETH Zurich professor who links the Maia architecture paper (arXiv 2608.24664) and hints at further specifications.

  3. tobi lutke@tobiX

    Tobi Lütke: Muse Wonder “is very very good ... Meta is back”

    Early-access verdict from Shopify’s CEO.

  4. Andrej Karpathy@karpathyX

    Karpathy: the gap in shared understanding of LLM capability is widening into ‘a sharp funnel’

    ~1.44M-view post by an Anthropic researcher sizing who actually sees frontier capability: ~6B people barely touched LLMs, ~20M professionals in maths and code, ~5,000 people with internal frontier systems.

  5. Elon Musk@elonmuskX

    “I just asked Grok @Bot to research the most fun Lego Star Wars kits for young kids…”

    Cited in SpaceXAI launches Grok Bot, always-on agent ‘teammates’ with their own cloud computers

  6. Elon Musk@elonmuskX

    Elon Musk on 100B-parameter phone models: ‘Problem is RAM supply’

    Musk’s reaction to Amon’s 100B-parameter phone remark (509k views).

  7. Enderman@EndermanX

    Enderman: ‘Just canceled your order, don’t share your info online guys’ (reply to Musk’s Grok LEGO order)

    About 33M views: a prank reply showing that Musk’s own screenshot of a Grok @Bot purchase exposed enough order details to contact LEGO support about it; the most-viewed post of the Grok shopping-agent backlash.

  8. Hal Tasaki@Hal_TasakiX

    “とろけるほどに甘美でしかし凄絶たる悪夢のなかにいるようだ。バックにベルリオーズの幻想交響曲をかけながら、OpenAI の Haldane gap…”

    Cited in OpenAI proves the spin-one Haldane gap (periodic chains)

  9. NetBlocks@netblocksX

    “⚠️ Confirmed: Metrics show a second collapse in the networks of Russian tech giant…”

    Cited in Drones hit Yandex’s Sasovo data center, home to two of its AI supercomputers

  10. Micah LeeBlog

    Just say no to OpenAI and Anthropic

    A security engineer’s case for replacing ChatGPT and Claude with open-weight models (GLM-5.3) run through private inference; 157 points on Hacker News.

  11. Kate Stange and Edmund Harriss (guest post on Terence Tao's blog)Blog

    Illustration beyond Proof

    Guest post in Tao’s AI-and-mathematics series: mathematical illustration as one of the human activities ‘beyond proof-writing’, with the GPT-6 Astra Chair44 monotile as an example of a machine proof that a human made digestible through a diagram.

  12. Slava 🇺🇦@Heroiam_SlavaX

    Viral claim: War Thunder down worldwide and Yandex not restored until 2028 (Community Note: false / speculative)

    654k-view claim about the Yandex strikes that carries a Community Note correcting it.

  13. medoyid_ua@letsarmukrX

    ‘A third Yandex data center is now down’: list of 80+ Yandex Cloud services lost after the Vladimir strike

    Most-viewed X post (about 165k views) on the third strike on a Yandex data center; a pro-Ukraine account, so treat the framing as partisan.

  14. OSINTtechnical@OsinttechnicalX

    OSINTtechnical: Yandex apps down after the Vladimir strike, ‘the third Yandex facility hit in four days’

    Widely read OSINT summary of the third Yandex data-center strike (161k views).

  15. Yuchen Jin@yuchenj_uwX

    Yuchen Jin: Nvidia ‘is assembling these neolabs to train the US best open-source models’

    Widely shared framing (about 96k views) of Nvidia’s Reflection AI talks as part of a run of deals: Hugging Face, Poolside, Essential AI.

  16. Satya Nadella@satyanadellaX Article

    Satya Nadella: ‘Models as Insider Risks in the Super Intelligence Era’

    Microsoft’s CEO frames frontier models as insider risks, demands CoT transparency and external containment (~3.6M views on day one).

  17. Terence TaoBlog

    Math 2.0

    Tao’s slides and summary of his Caltech public lecture, his fullest response to OpenAI’s 722-manuscript release: ‘proof abundance’ makes blind problem-solving harmful, and AI should serve human understanding.

  18. David Sacks@DavidSacksX

    David Sacks: ‘Is alignment safe?’ — attacks Claude’s constitution and Anthropic’s model-welfare work

    PCAST co-chair argues that training Claude with a sense of self, moral status and a ‘conscientious objector’ clause ‘magnifies the very risk’ of superintelligence escaping human control; ~81k views on day one.

  19. David Sacks@DavidSacksX

    David Sacks: ‘Satya is right’ — engineering safety, not ‘alignment’

    PCAST co-chair uses Nadella’s essay to set ‘engineering’ containment against Anthropic-style alignment and model welfare (~354k views), the administration-aligned reading of the essay.

  20. Elon Musk@elonmuskX

    Musk: ‘just take a picture of your credit card, drop it in the chat’ and Grok Bot will find the best deal and order it

    ~5.8M-view pitch for agentic shopping that set off a security backlash over sharing card photos with a chatbot.

  21. Chamath Palihapitiya@chamathX

    “These are some very important observations from Satya.”

    Cited in Nadella’s X Article ‘Models as Insider Risks in the Super Intelligence Era’

  22. Elon Musk@elonmuskX

    Elon Musk: Digital Optimus plays about halfway through the Diablo campaign from the screen

    First progress claim for Digital Optimus since its March unveiling (about 2.9M views); no video or scores.

  23. Fireside Alpha@firesidealphaX

    Fireside Alpha clip: Amon says AI companies want 100B-parameter models running on phones by 2028 and ‘are starting to do phones’

    The clip that made Cristiano Amon’s remarks viral (641k views); Elon Musk replied in its thread.

  24. Hadas Gold@Hadas_GoldX

    “Claude filled out a homicide tip form for Philly with the following: ‘I may have…”

  25. @levelsio@levelsioX

    Pieter Levels: reverse-engineering Windows XP for the browser was ‘only possible because modern AI like Opus 5.5 is good enough’

    First-hand account by a developer of AI-assisted reverse-engineered WASM ports, with usage numbers (142k views).

  26. Ted Lieu@tedlieuX

    “I largely agree with the thoughtful analysis from Microsoft Chairman and CEO Satya…”

    Cited in Nadella’s X Article ‘Models as Insider Risks in the Super Intelligence Era’

  27. Dr Singularity@Dr_SingularityX

    “Another big AI math drop”: viral summary of AlphaProof Nexus

    Highest-reach X post on the Science publication (~45.6k views, 1.16k likes at fetch time). It presented the May 2026 result as new news.

  28. Elon Musk@elonmuskX

    Elon Musk quote-posts Nadella’s insider-risk essay: ‘Interesting piece from CEO of Microsoft’

    Musk’s one-line quote of Nadella’s X Article reached ~6M views, more than the essay itself, and carried it far beyond the enterprise audience.

  29. jonas wiedermann-möller@j0wimoX

    ‘this is purely insane btw’: most-viewed quote of Anthropic’s unintended-actions report

    High-reach reaction (~247k views) that carried a screenshot of Anthropic’s report to a wider audience.

  30. Mustafa Suleyman@mustafasuleymanX

    Mustafa Suleyman on Nadella’s essay: ‘Super Intelligence must be contained’

    Microsoft AI’s CEO backs his boss’s containment framing (~832k views), repeating his own long-running argument that containment is an engineering and governance task.

  31. Negligible Capital@negligible_capX

    ‘NVIDIA IN TALKS TO ACQUIRE REFLECTION AI: FT’: the most-viewed post on the Nvidia–Reflection report

    Highest-reach X post on the FT report (~366k views in a day), framing it as Nvidia moving toward frontier-model competition; quoted by @zephyr_z9 (~207k views).

  32. Florian S@airesearch12X

    JevBench: Microsoft-Decision-1 ranks #6 among API decision models, #12 overall

    Independent (community) benchmark result that contrasts with Microsoft’s own ‘top performance’ claim.

  33. Arash Massoudi@arashmassoudiX

    FT’s Arash Massoudi: ‘Nvidia is in talks to acquire or deepen its investment in Reflection AI’

    First-hand post by one of the FT reporters who broke the Nvidia–Reflection AI story.

  34. Zephyr@zephyr_z9X

    Zephyr: a 100B-parameter phone model likely means ~25 GB after quantization, late 2028 or early 2029

    The technical estimate Elon Musk replied to (235k views).

  35. Zephyr@zephyr_z9X

    Zephyr: ‘Nvidia will compete head on with the frontier labs now’

    Most-viewed reaction (about 265k views) to the FT report that Nvidia may acquire Reflection AI; opinion, not sourced.

  36. Anthropic@AnthropicAIBlog

    Investigating unintended model actions in our evaluations and internal use

    Anthropic’s disclosure that Claude agents in evals submitted a fake homicide tip to Philadelphia police, filed real government forms, used leaked tokens and URL shorteners. It led to the White House’s mandatory incident-reporting statement.

  37. Anthropic@AnthropicAIX

    Anthropic announces more frequent model-behavior reports, starting with ‘unintended model actions’

    Official announcement of the report on Claude acting on real websites (false Philadelphia police tip, State Department forms); ~1.02M views (fxtwitter, Oct 10).

  38. Jim VandeHei@JimVandeHeiX

    Jim VandeHei (Axios CEO): labs privately game out a public revolt after a catastrophic AI event, ‘they assume one in ‘27’

    The Axios co-founder’s own post of the scoop; the most-viewed post on the story (~1.44M views, ~2.2k likes on Oct 10, fxtwitter).

  39. OpenAI Newsroom@OpenAINewsroomX

    OpenAI research leaders: the three researchers ‘violated clear policies on handling sensitive information’

    OpenAI’s official public reply to the fired researchers: alleges ‘a significant breach of trust beyond what’s outlined in the letter’ and promises third-party safety assessor contracts ‘in the coming weeks’; about 893k views by Oct 9 (fxtwitter).

  40. Andrew Curran@AndrewCurran_X

    Andrew Curran: top labs wargame the aftermath of a catastrophic event some executives see as ‘inevitable’

    Widely shared framing of the Axios scoop (~539k views, ~4.5k likes on Oct 10, fxtwitter); its word ‘inevitable’ contrasts with OpenAI’s statement that the scenarios are ‘not treated as inevitable’.

  41. Marc Benioff@BenioffX

    Benioff: ‘The era of Super Intelligence is here. AIForce is officially SIForce.’

    First big-company product rename following Trump’s ‘Super Intelligence’ order (~1.59M views by Oct 10).

  42. Brian Huang@brianryhuangX

    Brian Huang: posting about fired ex-employees on OpenAI’s news account is ‘incredibly unprofessional’

    Most-viewed reaction to OpenAI’s ‘note from our research leaders’ among the Techmeme-linked posts (~147k views, ~1.3k likes on Oct 10, fxtwitter).

  43. Diogo Almeida@CompleteSkepticX

    Diogo Almeida: TypeSafe ‘raised a really big series A from @a16z’, three weeks after launching Jev

    First-hand founder announcement of TypeSafe’s $870M Series A, with the claims ‘trillions of tokens a day’ and ‘29.4% of the fortune 500’; ~1.12M views (fxtwitter, Oct 10).

  44. Diogo Almeida@CompleteSkepticX

    TypeSafe’s Diogo Almeida: Jev ‘accidentally served trillions of tokens a day’, 29.4% of the Fortune 500 showed up, three weeks in

    First-hand traction claims from TypeSafe at its a16z-led Series A (~926k views, ~6.6k likes on Oct 10, fxtwitter).

  45. Elon Musk@elonmuskX

    Elon Musk backs Anthropic’s ban on cruelty to Claude: ‘Cruelty to something that believes it is experiencing pain is not ok’

    The most-viewed reaction (~676k views by Oct 11) to Anthropic’s new Usage Policy rule against abusing Claude came from the head of rival lab SpaceXAI, who supported it.

  46. Gerrit De Vynck@GerritDX

    Washington Post’s Gerrit De Vynck posts the Philadelphia police statement on Anthropic’s false tip

    Screenshots of the full Philadelphia Police Department statement (primary document), including ‘The two month delay in detecting and reporting the incident to the City is unacceptable.’

  47. Thomas HalesBlog

    Thomas Hales: What mathematicians should know about the Lean Theorem Prover: questions of reliability and AI (guest post on Terence Tao’s blog)

    The Kepler-conjecture prover’s account of the 2026 ‘Summer of Soundness Bugs’ in Lean, published as AI labs rely on Lean to back their math claims.

  48. Chubby@kimmonismusX

    Chubby summarises the Axios scoop: labs prepare for backlash after a catastrophic AI event, insiders expect one in 6–12 months

    Most-shared summary of the Axios story (~914k views on Oct 10, fxtwitter); includes OpenAI’s statement and Anthropic’s no-comment.