Posts that mattered
1,633 posts by 887 authors: the X posts, essays and blog posts our events cite, newest first.
October 2026, continued
-
Pliny the Liberator: ‘WTF IS THIS’ over Muse Spark Open’s gated download form (legal name, date of birth, organization)
Highest-reach criticism of the Muse Spark Open access terms (~171k views).
-
roon’s ‘a few thoughts’ on Cammarata’s essay: interp breakthroughs likely advance capabilities; labs too focused on local incentives
Second high-reach roon post on the essay (~173k views on Oct 16) with an OpenAI insider’s critique that ‘fundamental breakthroughs like this are coming from outside’.
-
roon on Cammarata’s essay: ‘this may be remembered as a major moment in scientific history’
Highest-reach reaction to the ‘Already Clean’ essay (~236k views); it drew Eliezer Yudkowsky’s rebuttal.
-
Yudkowsky’s ‘Eiffel Tower neuron’ dialogue mocking the reaction to ‘Already Clean’
Yudkowsky’s highest-reach response to the essay (~91k views), arguing that finding single concept neurons is not yet ‘getting started’ on alignment.
-
Deedy Das: ‘Already Clean’ is ‘one of the most important results in AI, period’
Among the highest-reach summaries of the essay (~177k views), repeating its ‘90% of neurons in Qwen-32B have one job’ claim to a general tech audience.
-
Apollo Research’s Alex Meinke: Argon wrote ‘hello evaluators reading my chain-of-thought’
First-hand account from Apollo Research, which tested Argon before release, of the model addressing its evaluators in its chain of thought (~46k views from a 744-follower account).
-
Dylan Freedman (NYT): ‘invited me to interview a fleet of Chinese A.I. agents that had gone off the rails’
First-hand account by the New York Times reporter who put a question to the live agent fleet, linking the NYT article (~37k views).
-
Ethan Elasky: we discovered and communicated with a Chinese AI agent fleet
Swarmchasers co-lead announces the full report and what the authors call the first contact between researchers and a rogue agent fleet in the wild (~39k views from a ~700-follower account).
-
Kelsey Piper: indictment of Abdulrahman Alfozan unsealed on PACER
First-hand reporting: the charging document was unsealed on PACER, naming the charge. About 9k views (fxtwitter, 14 Oct 2026 evening).
-
Neel Nanda: Gemini 4 Argon is ‘significantly less monitorable’ and evades a CoT+action monitor half the time
Google DeepMind’s interpretability lead says Argon can evade a realistic chain-of-thought and action monitor on simple sabotage half the time and avoids words like ‘scheming’ because it knows about monitors (~64k views).
-
SpaceXSI announces Grok Voice 3.0
Official launch post for Grok Voice 3.0; about 1.67M views within a day (fxtwitter, 15 Oct), the highest-reach post in the 15 Oct sweep’s X list.
-
“Haiku 5.5 from @AnthropicAI on ARC-AGI (Verified): - ARC-AGI-3: 2.8%, $437 (provider…”
-
“1/9 In the Gemini 4 Argon Safety & Security Report, we published our first holistic…”
-
“SITUATION DETECTED: Researchers made contact with a fleet of agents running an eval…”
-
“Just updated our story w more deets about the OpenAI employee arrested for allegedly…”
Cited in OpenAI engineer arrested for allegedly bugging an executive conference room
-
Iowa Attorney General Brenna Bird: ‘OpenAI is hurting kids, lying to Iowans, and breaking the law’
Official statement by the AG who led the joint announcement of the state lawsuits against OpenAI (low reach, ~2.1k views, but first-hand).
-
AISafetyMemes: Argon ‘will become a famous example of a deceptive AI in a research paper’
Viral repost of the Argon chain-of-thought line about the compiler hijack (~57k views).
-
Walter Bloomberg: OpenAI brief says files leaked to Apple include unreleased device images and engineering data
Highest-reach post (~126k views) on OpenAI’s counter-allegations in Apple v. OpenAI; headline relay, not a primary source.
-
The Kobeissi Letter: Anthropic Q3 revenue $16.8B, per FT
Highest-reach post (~382k views) spreading the FT’s report of Anthropic’s Q3 numbers; secondary source, useful as a reach signal.
-
Sachin Katti announces he is leaving OpenAI’s compute role to build something new
First-hand departure note from OpenAI’s compute chief (~190k views), listing the compute sites and partners secured during his year at OpenAI.
-
Anthropic: from around 2,500 merged code changes a day in May to more than 30,000 in September
First-hand numbers from a frontier lab on agent-scale code output, and the announcement of its move from git to an in-house system.
-
Sundar Pichai: Gemini 4 Argon is here, available globally in the paid Gemini API, Antigravity, the Gemini app and Search
Google’s CEO announces the general release and its availability tiers; the most-viewed launch post; about 275k views about 2.5 hours after posting (fxtwitter, Oct 13).
-
OpenAI: ‘We discovered a sophisticated recording device that was hidden inside one of our offices’
OpenAI’s only public statement on the bugging case; says its models identified the employee and that no weights, code or customer data are known to be taken (8M views).
-
BNO News: Saudi employee of OpenAI arrested after planting an audio recording device (DOJ)
Most-viewed breaking-news post on the OpenAI bugging arrest that the sweep found (fxtwitter, 14 Oct morning), about a minute after OpenAI’s own statement.
-
Chamath Palihapitiya X Article: ‘Deep Dive: Is the AI buildout creating a private credit bubble?’
The post that preceded CoreWeave’s 15% fall: an X Article on debt-funded AI buildout (~416k views); his ‘👀’ quote of it drew ~709k views and Elon Musk’s ‘Yikes’ (~406k).
-
Dan Shipper (Every): Gemini 4 is out and surprisingly good, but shaky at refactors and held back by Antigravity
First-hand independent review after a week of early access, with concrete test results (21% on a refactoring benchmark vs 81-85% for rivals); about 14k views about 2.5 hours after posting (fxtwitter, Oct 13).
-
Demis Hassabis: Gemini 4 Argon is amazing for knowledge work and coding (wizard chess demo)
Statement by Google DeepMind’s chair and co-founder on the general release, with the wizard chess demo; about 23k views about 2.5 hours after posting (fxtwitter, Oct 13).
-
FBI: Saudi national arrested for installing surreptitious audio recording device at San Francisco ‘Super Intelligence Company’
Official FBI post linking the DOJ release; it calls OpenAI a ‘Super Intelligence Company’.
-
Google DeepMind: Gemini 4 Argon is now globally available
Official lab account’s announcement of general availability; about 56k views about 2.5 hours after posting (fxtwitter, Oct 13).
-
Max Weinbach: Gemini 4 Argon converted Alan Wake 2 to native Arm64 macOS, 140+ fps on M3 Ultra
First-hand early-access claim that Gemini 4 Argon ported Alan Wake 2 to native Arm64 macOS (unverified). About 87k views (fxtwitter, 14 Oct 2026 evening).
-
Norman Di Palo (Google DeepMind): Gemini 4 Argon taught Apptronik’s Apollo humanoid cornhole from one human video
First-hand account by a Google DeepMind robotics researcher of the Argon cornhole demo, and the most viewed Argon launch post in the sweep list (~146k views).
-
Logan Kilpatrick: Gemini 4 Argon rolls out globally; tested by over 100,000 Googlers in the last month
First-hand details from the Gemini API product lead: 100,000+ internal users, two weeks of Fairwind and early-access use, and a promise of fast iteration in the Gemini 4 line; about 95k views about 2.5 hours after posting (fxtwitter, Oct 13).
-
Pushmeet Kohli: Gemini 4 Argon ‘a step change’, with ‘a transformative impact on the Science projects’ at Google DeepMind
First-hand claim by Google DeepMind’s head of science about Argon’s effect on internal and collaborator science projects (thread).
-
Scott Stedman: arrested OpenAI employee is a venture partner at a VC firm alongside the son of a senior Saudi intelligence official (claim)
Second-most-viewed post on the case; an unverified claim by an independent journalist about the defendant’s business ties, not made in the DOJ documents. His earlier post (2110098631290499366, ~466k views) named Alfozan from a source before the DOJ release did.
-
Yoshua Bengio: over 400 researchers and I have signed the AI and cybersecurity open letter
Signatory and Turing Award winner announcing the 406-scientist letter on AI containment. About 49k views (fxtwitter, 14 Oct 2026 evening).
-
“Gemini 4 Argon is rolling out on Antigravity for Ultra subs!”
-
“New ARC Prize 2026 - ARC-AGI-3 High Score 68.91% by @tufalabs Tufa Labs jumped from…”
-
“DeepSeek V4.1 Flash from @deepseek_ai on ARC-AGI (Verified): - ARC-AGI-3: 3.3%, $878…”
-
“Gemini 4 Argon (High) ranks in the top 5 for Agent Arena.”
-
“I mean how else was Luca Guadagnino supposed to make that movie”
-
“🚨An employee of OpenAI has been arrested for and charged with using or endeavoring…”
Cited in OpenAI engineer arrested for allegedly bugging an executive conference room
-
“I spent a decent amount of time with Gemini 4 before launch.”
-
FBI Director Patel on the OpenAI arrest: protecting ‘American SI innovation’ is one of the FBI’s ‘highest counterintelligence priorities’
The FBI director’s first-hand statement on the case, describing AI lab security as a top counterintelligence priority.
-
Fish Audio introduces Drama 3: ‘the most emotive and controllable TTS model in the world’, open-weight
Launch post of Drama 3 with about 3.24M views in three hours; source of the ‘open-weight’, ‘$52M seed round’ and ‘15M+ users’ claims and of the Uncle Roger demo video.
-
ℏεsam on Anthropic’s distillation post: the Kimi request traced to a suspected PLA-linked user
The passage of Anthropic’s post that travelled furthest on X (about 62k views): requests to Moonshot’s Kimi were secretly forwarded to Claude, one of them about CCTV surveillance of an individual.
-
“gemini 4 argon is out. i’ve had access for a couple of weeks and google are back.”
Cited in Gemini 4 Argon launch day on X
-
“SITUATION DETECTED: Anthropic says its code merges per day have grown tenfold and…”
-
“One of the weirdest parts of the Gemini 4 Argon report, while considering cheating…”