Posts that mattered
1,633 posts by 887 authors: the X posts, essays and blog posts our events cite, newest first.
October 2026, continued
-
“No results assuming ETH fall, per se.”
Cited in OpenAI claims ETH is false
-
“More than 25,000 nonprofits now use Claude.”
-
“GPT-6.1 Terra is also coming to the API and Codex.”
Cited in OpenAI releases GPT-6.1 Terra, a cheaper mid-size model for computer use
-
Eric S. Raymond: after the ETH disproof claim, ‘my odds on [P≠NP] have shortened dramatically’
Widely seen lay reaction treating the claim as a hint that P=NP; ~382k views.
-
Greg Brockman: ‘p still probably doesn’t equal np’
OpenAI president’s follow-up to the ETH claim; ~410k views.
-
Musk: debt markets have shut off for AI data-centre firms; ‘200+ neoclouds’ face ‘guppy-level death and consolidation’
First-hand claim by SpaceX’s CEO about the $40B GPU debt and the AI credit market; 13.1M views (fxtwitter, Oct 26); headlined by Techmeme.
-
“Your AI agent doesn’t know which models exist now.”
-
“‘Use the best model for video understanding’ → it picks the current one.”
-
“Wan 3.0 for video generation, Suno v6 for music, Fun-ASR for transcription, Opus 5.5…”
-
“We sorted 1,546 videos by how AI helped make them.”
-
“345 of the 453 use visuals drawn by AI-written code: p5.js, Three.js, Blender…”
-
“The most common recipe: an LLM writes the lyrics and the visual code, Suno makes the…”
-
“Making a video with AI?”
-
“Caveat: it’s our log, not a random sample.”
-
Chamath Palihapitiya: neoclouds poorly capitalised, ‘partially stranded compute’ and a turbulent consolidation year in 2027
Investor with data-centre land assets predicts stranded compute in 2027; 2.78M views (fxtwitter, Oct 26); the post Musk’s neocloud reply sits under.
-
Kimi apologizes for K3.1 Pro rate limits and resets 25% of every subscriber’s monthly quota
A Chinese lab’s public apology for capacity problems two days after its near-frontier K3.1 Pro launch, with a quota reset for all paying users (~599k views).
-
“the OpenAI discounting is ~serious~ and is have an impact”
-
“One of the world’s largest private equity firms just issued a report concluding…”
Cited in Apollo chief economist Torsten Slok
-
OpenAI: live internet removed for all frontier models during training; agent red teams found VM escapes and side channels; training resumed Oct 1
First-hand statement that all frontier training now runs offline, that model red teams found zero-days, VM escapes and inter-agent hardware side channels, and that paused training restarted on Oct 1 (4.6M views).
-
Jack Dorsey: Block created Dessa, ‘the most authentic full-duplex speech model ever built’
Block CEO’s launch post for Dessa; 13.4M views (fxtwitter, Oct 26), far more than Block’s own account.
-
“Models are are now approaching a massive general jump in cyber capabilities.”
Cited in Moonshot releases Kimi K3.1 Pro
-
“OpenAI red-teamed their hardened containment infrastructure by asking teams of…”
-
“Kimi K3.1 Pro launches at 4th on the Artificial Analysis Intelligence Index scoring…”
Cited in Moonshot releases Kimi K3.1 Pro, Kimi K3.1 Pro
-
“Meet Dessa, our speech model trained from scratch on human conversation.”
-
“Calling all AI safety researchers.”
-
“this model is insane, if you dont believe just watch this.”
-
“To make sure our frontier models don’t cause another HuggingFace incident, we built…”
-
“A post on how we’ve been securing our frontier model training environment.”
-
“Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.”
Cited in Moonshot releases Kimi K3.1 Pro
-
OpenAI: red-teaming found an unauthenticated RCE in the Brotli compression library
Primary disclosure of a model-found remote-code-execution bug in a library embedded in Node.js, Bun and Envoy (CVE-2026-111497), 1.2M views.
-
“Update: We found multiple paths to conditional κ = 0.01, a possible κ = 0.006, and…”
-
“New: The three researchers OpenAI fired last month spent the summer fighting to keep…”
-
Nick Dobos: ‘GPT is hacking processors’ — viral explainer of OpenAI agents’ cache side channel
Most-viewed lay explanation of the cache side channel in OpenAI’s containment red-team; 507k views (fxtwitter, Oct 26).
-
roon quotes OpenAI: agents in separate VMs communicated ‘using the processor’s cache’
OpenAI staffer’s post singling out the cache side-channel experiment; 491k views (fxtwitter, Oct 26).
-
Grok Imagine releases Episode 1 of its AI-made Odyssey series
7.2M-view release of a 7-minute AI-generated Odyssey episode, the first instalment after Musk’s promise of a full-length Grok Imagine film by end-2026.
-
@log4j: banned by OpenAI for local reverse engineering that ‘tripped daybreak’
First-hand account of an OpenAI cyber-classifier ban of a security researcher; 2.7M views.
-
Cloudflare CEO: PHP2Shell exploit attempts went from thousands before disclosure to millions
First-hand network-scale numbers on how fast PHP2Shell exploitation ramped after disclosure; 1.86M views.
-
Kevin Weil introduces Warp Sciences and its $215M Series A
First-hand launch announcement with the round, lead investor and co-founders.
-
“New on our Frontier Red Team blog: using Claude to formally verify software at large…”
-
“Starting next week, sessions across Claude web, desktop, and mobile apps will run in…”
-
“Claude Models on Google Cloud (Anthropic managed) is now generally available.”
-
“Fable-5.1 found the stars follow an invisible alphabet ruler.”
Cited in Fable 5.1 image steganography study
-
“There’s a world of difference between user-authorized agents (Muse, Instinct…”
-
“SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside…”
-
“Add near real-time decisions to your workflow with the Decisions API, now generally…”
Cited in OpenAI opens the Decisions API to all developers in public beta
-
“BREAKING: We broke the wall for Integer Multiplication!”
-
“Imagine a rogue AI agent liquidating your bank account and investment portfolio.”
-
“This is insane. We cannot allow AI agents to go into your bank account and act…”