Sunday, August 2, 2026

The Bugged Dev chained GPT Image 2.0, TripoAI, and OpenAI Codex into a pipeline that turned one design image into a rotatable, hotspot-annotated 3D human anatomy app — then used Codex to cut total asset size by 97%.

FFmpeg's official account thanked Cursor for providing free AI coding credits to several of its developers. The gesture drew mostly praise from a community that depends on FFmpeg as invisible infrastructure — plus a few jokes about AI companies eyeing critical open-source projects.

Tokens explain consumption; dollars explain budget impact. Cursor’s quick reversal restored both the Usage-page Cost column and historical CSV values, but the dispute exposes a broader problem in agent pricing.

Fable-OS is more substantial than an “AI operating system” webpage: it is a from-scratch x86_64 kernel whose primary interface is an agent. We inspected the source, ran its 39 host test suites, and separated the impressive audio-driver result from the physical-hardware, security, cost, and licensing caveats.

Solving problems that resisted mathematicians is a major capability signal. It still does not meet the classic broad definition of superintelligence. The useful concept in between is jagged, domain-superhuman intelligence.

Opus 5 ran for roughly two hours and wrote 5,500 lines of procedural Three.js code from one literary paragraph. Karpathy’s follow-through point: LLMs have the stamina to build hyper-custom worlds nobody would hand-author — but still struggle to watch, play, and audit them.

The study’s “surprisingly good” result means LLM recommendations moved simulated households toward life-cycle theory. It does not mean chatbots beat advisers, predict markets, or safely replace individualized financial care.

Astra arrived as a research dossier rather than a benchmark chart. OpenAI named its next major model family, showed ten claimed frontier results, and left pricing, access, architecture, and product dates unannounced.

Astra’s results range from non-sofic groups and sphere packing to quantum games and circuit lower bounds. The Lean files make this unusually auditable, but machine checking is not the same as complete community validation.

A July 27, 2026 X demo from @mikeluan123 turned Matt Shumer's viral "one-shot AAA" prompt template into a Homeworld-style universe-scale space RTS running on Claude Opus 5 and Three.js — no art assets, fan-out sub-agents, and a published cost breakdown that split reactions between awe and disbelief.

Hours after Claude Opus 5 launched, am.will shipped a browser Rocket Arena clone with reflections, Rapier physics, and bots — then filmed Opus playing the game itself. explainx.ai breaks down the demo, stack, and what it means.

Three vendors, one Thursday: @ClaudeDevs refills Claude buckets, Theo flags a Codex reset hours later, and @leerob doubles Cursor model quota fleet-wide. explainx.ai explains 5-hour vs weekly vs banked resets — and why Fable churn pressure keeps the arms race hot.

Training aligned actions failed. Teaching Claude why — constitution, difficult advice, fictional stories — fixed agentic blackmail OOD. explainx.ai explains the 3M-token dataset win and what agent builders should copy.

Holaboss’s holaOS repo frames agents as needing an OS-like layer: runtime, tools, apps, and state that survives across runs. Here is a builder’s map of the thesis, the stack, and the Reddit conversation around “role-holding” workspaces.