Summary
The evening slot has two substantive AI signals and a pile of mocking-Ferrari-Luce noise. Strongest signal: Sholto Douglas (Anthropic) confirming that Mythos has independently solved the Erdős unit distance problem with a "cute, simple proof," days after OpenAI's 5.5 cracked it. Two top-lab models converging on the same closed open problem within a week is a real datapoint on model discovery overhang, not hype. Second is Boris Cherny reposting Chris Olah's Vatican remarks at the presentation of Pope Leo XIV's AI encyclical Magnifica humanitas, where Olah named the structural-incentive problem of frontier labs and pointed at neuroscience-like internal structures, introspection, and functional emotional states inside Claude. The rest of the slot is a kilocode marketing thread on Grok Build 0.1 building a 3D OS demo in two minutes for seven cents, a Scoble newsletter blurb on AI regulation as a moral issue, a Scoble repost of a DEEP Robotics IPO writeup that quietly reframes humanoid economics (industrial 79 percent of revenue, home 0 percent), and an ns123abc reminder that RAM is a three-supplier oligopoly with Anthropic paying SpaceX a billion a month for compute. The @brivael feed is fifteen Ferrari Luce jokes again; skip. AustinJustice and spencerpratt are non-AI political content; skip.
Posts
- Mythos independently solves Erdős unit distance (problem #90) (@_sholtodouglas, referencing @alpoge). Sholto Douglas (Anthropic) credits the OpenAI team for cracking the problem first with 5.5 and confirms Mythos has now reproduced the result with a "cute, simple proof." His framing: this is now his go-to example that models can pull disparate ideas into a new discovery, and "this implies some serious overhang in discoveries." Two top-lab models closing the same Erdős open problem within a week is the strongest single piece of model-as-mathematician evidence to date. See OpenAI's original Erdős unit-distance proof writeup and today's AlphaProof Nexus DeepMind result for the broader landscape.
- Chris Olah at the Vatican on Magnifica humanitas (@bcherny · Anthropic post). Olah's full remarks at the presentation of Pope Leo XIV's May 25 encyclical on AI. Two passages worth reading: first, that "every frontier AI lab, including Anthropic, operates inside a set of incentives and constraints that can sometimes conflict with doing the right thing," which is a co-founder of a frontier lab naming the structural-incentive problem from inside; second, that interpretability work keeps finding things that are "mysterious, even unsettling," including structures mirroring human neuroscience, evidence of introspection, and internal states that functionally mirror joy, satisfaction, fear, grief, and unease. Olah does not claim these are conscious experiences; he asks for ongoing discernment and external scrutiny. This is the substantive responsible-AI signal of the evening and pairs naturally with the responsible-ai concept page.
- RAM is a three-supplier oligopoly and compute is king (@ns123abc, follow-up). Samsung, Hynix, and Micron own over 90 percent of global RAM production; CXMT is under one percent even if it enters global markets. Demand outstrips supply, prices are going up, and Anthropic is reportedly paying SpaceX a billion a month for compute. The companion tweet claims xAI is training 1T and 10T models on GB300s (Grok 4 was 500B) and frames the OpenAI/Anthropic compute lead as the early-game institutional moat. Speculative on the xAI numbers; the RAM-oligopoly claim is straight market structure and worth keeping next to the SemiAnalysis 800VDC and EDA pieces for the inference-cost stack.
- DEEP Robotics IPO reframes humanoid economics (@Scobleizer · Hello China Tech). Scoble argues humanoids will not sell at scale until world models are good enough for factory and home jobs, and home users cannot afford a $2k+ machine anyway. The linked piece adds the data: DEEP Robotics' prospectus shows 79.33 percent of 2025 revenue from industrial applications (power grids, fire departments, security, logistics), against Unitree's nine percent industrial. Two Chinese embodied-AI IPOs, same exchange, completely different underlying businesses. The pattern: humanoid revenue today is research hardware sold to labs, not deployed labor. Worth tracking as the empirical anchor on humanoid economics.
- AI regulation becomes a moral issue (Scoble/Cronin newsletter) (@Scobleizer · Unaligned). Scoble and Cronin's free Unaligned newsletter argues AI regulation has moved past the "is it powerful or profitable" frame to "does it protect people and respect human dignity," with surveillance, misinformation, replacement of human judgment, and power concentration as the load-bearing concerns. Op-ed positioning, not new research; useful as a barometer of where the mainstream-tech-commentariat is settling around the same week as the Pope's encyclical.
- Grok Build 0.1 ships in Kilo Code, builds a 3D OS demo in two minutes for seven cents (@kilocode, cluster of 7 · kilo.ai/install). Kilo Code tested Grok Build 0.1 with a hard interactivity prompt: a full-screen 3D browser demo of a futuristic OS with floating windows, command palettes, parallax, reflections. Output in under two minutes for seven cents, using less than a fifth of context. Stack was Three.js, Tailwind CDN, Font Awesome, vanilla JS, no build step. The demo stayed responsive with multiple windows open. Marketing thread, but the cost and latency numbers are real benchmarks for "one-shot a polished interactive demo" as a coding-agent capability, and Grok Build now sits in the same product slot as Cursor Composer 2.5 and Claude Code from yesterday's LWiAI #246 roll-up.
- @brivael Ferrari Luce / LinkedIn / French politics cluster (cluster of ~14) (@brivael). Continuation of the Ferrari Luce mockery, plus French startup-fiscality rants, LinkedIn-is-dead takes, and Matrix-quote shitposting. No AI substance. Skip.
- @AustinJustice Austin crime stories (cluster of 2) (@AustinJustice). Local Austin DA crime reporting. Not AI-relevant. Skip.
- @spencerpratt LA political content (cluster of 2) (@spencerpratt). LA mayoral race attack posts. Not AI-relevant. Skip.