Posts that mattered
743 posts by 495 authors: the X posts, essays and blog posts our events cite, newest first.
September 2026, continued
-
“I had Opus 5.5 recreate the city of San Francisco in Unreal Engine.”
Cited in Opus 5.5 recreates San Francisco in Unreal Engine, with everything powered by Jev
-
“Claude Opus 5.5 1-shotted the launch video of @shotbaseapp using @HyperFrames_ in…”
Cited in Opus 5.5 “1-shotted” the Shotbase launch video with HyperFrames in under 20 minutes
-
“Opus 5.5, The Minecraft Test.”
Cited in “Opus 5.5, The Minecraft Test”: browser Minecraft clone with sunset shaders
-
Source for those who were asking (PDoomVideo repo)
Links the open-source code github.com/JohnHeibel/PDoomVideo.
-
“sketch-to-simulation with Opus 5.5”
Cited in “Sketch-to-simulation with Opus 5.5”: pencil-sketch catapult turned into a 3D physics sim
-
“opus 5.5、本当にやばすぎる... 5分もかからずこのピクセルアニメーション出してきやがった。 プロンプトをリプ欄に貼っておく👇”
Cited in Pixel-art critter runs along a cosmic rainbow road, made by Opus 5.5 in under 5 minutes
-
“Asked Claude Opus 5.5 to animate its own life, from day 0 to now No video model.”
Cited in “Asked Claude Opus 5.5 to animate its own life, from day 0 to now”
-
“First Opus 5.5 vs GPT-6 Astra test is 3D.”
Cited in Opus 5.5 vs GPT-6 Astra: one prompt, Blender only, all procedural
-
“From sketch to home, built in 3D with Claude Opus 5.5 using Three.js + TSL.”
-
“I had early access to Opus 5.5 and it blew my expectations clear out of the water!”
Cited in Early-access Opus 5.5 creations, starting with a Dark Souls-style game
-
“This is the first release in our open-source week: runtime dynamic compression and…”
Cited in Tim Dettmers’ dlab open-source week: 1.5-bit inference and a 125B model on one 24 GB GPU
-
“I asked Opus 5.5 to try a bunch of redesigns of my personal website using my MAX sub…”
Cited in Opus 5.5 redesigns a personal website in many iterations, then cuts a trailer of them
-
“holy shit, opus 5.5 is kind of insane at animation.”
Cited in “small print”
-
“Claude Opus 5.5 just ONE SHOTTED this Roblox Anime Super Smash Bros Game.”
Cited in Opus 5.5 “one-shots” a Roblox anime platform fighter
-
OpenAI announces priorities and principles for third-party assessments
Announcement of OpenAI’s principles for independent assessments of safety cases, safeguards and misalignment incidents.
-
POLITICO Europe: Burnham says UK-led AI agreement will seek global principles and standards
Press post reporting UK PM Andy Burnham’s plan for a UK-led global AI agreement.
-
‘Opus 5.5 benchmarks just dropped’
Viral launch-day benchmark reaction (157k views); claims unverified.
-
‘Opus 5.5 costs 40% less than Opus 5, performs at Fable 5.1 level’
Widely shared launch-day summary of Opus 5.5 (276k views).
-
‘Anthropic is now offering banked resets’
Most-viewed reaction (467k) to Anthropic’s saveable rate-limit reset announced with Opus 5.5.
-
Advisory Group on Mathematics and Artificial Intelligence
Source of OpenAI’s claim that an internal model resolved 100+ long-standing open math problems; creates an IAS-hosted review body.
-
Announcing the Advisory Group on Mathematics and Artificial Intelligence
Nine leading mathematicians (Gowers, Hairer, Witten, Vakil, Wood…) formed an unpaid, independent group to advise OpenAI on releasing its 100+ claimed math results.
-
Building standards for the next phase of AI
OpenAI says fully autonomous recursive self-improvement should not be pursued until safe and proposes international technical standards via safety institutes.
-
Toby Ord: ‘Swarm Scaling’ — how agent-swarm performance scales with swarm size (λ ≈ 0.48–0.68)
First quantitative estimate of the ‘parallelizability’ of AI agent swarms, from OpenAI’s GPT-5.6 swarm charts; Ord argues the measured λ keeps recursive self-improvement scenarios plausible. Featured in Import AI 475.
-
Xiaomi announces MiMo-V2.6 series
Launch post for the top open-weights model on Artificial Analysis at release.
-
Nat Friedman: Muse built from scratch but ‘heavily inspired’ by OpenClaw
First-hand statement from a Meta Superintelligence Labs leader on Muse’s origins (1.2M views).
-
“The community response to the Millennium Problems for Biology has been great.”
-
Amazon: $1.9B more for Delivery Service Partners; 20,000+ Smart Delivery Glasses by end of 2027
Official Amazon X Article (Ignite Live, Sept. 21) giving the smart-glasses deployment target and other AI delivery tools.
-
Peter Steinberger (OpenClaw) on Muse: ‘They built their own agent, being inspired’
OpenClaw’s creator clarifies Meta did not use OpenClaw code in Muse (578k views).
-
‘Muse is LITERALLY Openclaw for normies’
The viral comparison (990k views) that prompted Nat Friedman’s and Peter Steinberger’s replies about Muse and OpenClaw.
-
Qwen: Meet Qwen-Image-2.1, now open weights
Official launch post for Qwen-Image-2.1 (2.27M views).
-
“我用 ChatGPT pro （GPT 6 Astra）做了一个陶渊明的桃花源记，记得点右下角的‘循文入境’按钮听朗诵。 朗诵不是 AI…”
Cited in Peach Blossom Spring (桃花源记) as an interactive web scene, made with ChatGPT Pro (GPT-6 Astra)
-
‘An open-weight 7B model that outperforms Nano Banana 2.0’
High-reach community claim (563K views) that framed Qwen-Image-2.1’s reception; not an official benchmark.
-
Paul Graham on why people suspect ulterior motives when AI labs ask for regulation
Prominent reply (233K views) to Jensen Huang’s claim that labs seek regulation for ‘ulterior reasons’.
-
Sebastian Raschka: Jev’s breakthrough is that a classifier generalizes
Assessment by a well-known ML researcher (238k views).
-
Andrew Chen: Jev enables ad-supported AI-native apps
a16z partner’s thesis on Jev’s economics (102k views).
-
‘I have jev access and i genuinely can’t think of one thing to use it for’
Sceptical counterpoint to the Jev hype (167k views).
-
Trump polls followers on renaming AI ‘Superior’, ‘Extreme’ or ‘Supreme Intelligence’
Origin of the administration’s ‘SI’ renaming of AI; 15.6M views, the highest-reach AI post of the week.
-
Sam Rodriques launches the Millennium Problems for Biology (‘the last reasonable eval for AI in biology’)
First-hand launch of FutureHouse/Edison’s twelve lab-verifiable grand challenges for AI in biology (~508k views).
-
“Meet Qwen3.8-LiveTranslate, Qwen’s next-generation real-time simultaneous…”
Cited in Alibaba launches Qwen-Audio-3.1 five-model voice stack and cuts audio API prices up to 95%, Qwen3.8-LiveTranslate (Flash Realtime)
-
Full text of Trump’s Truth Social post announcing the ‘AI Force’ and an AI czar
Mirror of the primary Truth Social post (TS Sep 19 2026, 1:12 PM ET) cited by the AI Force entry; 84K views.
-
Why do we need human mathematicians anymore?
Guest essay proposing the axiom ‘We (humans) should help humanity flourish’; it tallies the mathematicians’ collective statements (Leiden Declaration 4,000+, Math and AI 7,000+, anti-Mathathon letter 2,000+).
-
Nimble: an open alternative to Jev
Points to Bespoke Labs’ open Nimble model as a Jev alternative (168k views).
-
Andrew Curran: Trump announces the US AI Force and a coming AI czar
Most-viewed relay of the AI Force announcement (582K views).
-
Realtime virtual try-on hauls built with Jev
High-reach Jev demo (559K views), built with TypeSafe for Drape.
-
Jev as a model router
Popular use-case post (168k views).
-
JevBench results
Community benchmark for Jev-class models (212k views).
-
‘Jev is INSANE’ — trading bot that lost $31,680
Most-viewed Jev user post (1.4M views); an ironic demo of an autonomous trading bot.
-
Justine Moore: Jev classifies thousands of Zillow listings for $0.18
Popular Jev demo (110k views).