今日精读

"We've reset usage limits across Codex and ChatGPT Work... We made it too easy to use the highest-compute settings without making the impact on usage limits sufficiently clear." / "GPT-5.6 Sol Ultra... produced a proof of the 50-year-old Cycle Double Cover Conjecture using 64 subagents in just under one hour."

"Grok 4.5 even ranks slightly above Fable... on some software benchmarks!" / "Grok 4.5 is Opus class for browser use"

"GPT-Live is now fully rolled out to all ChatGPT users globally. We're also doubling everyone's voice usage limit for the whole weekend"

"GPT-5.6 Luna outperforms GPT-5.5 at its highest reasoning setting while costing 25x less."

"We're inviting researchers... to try to find a universal jailbreak that can defeat our predefined biosafety challenge against OpenAI's frontier models."

"With GitHub Issue Fields, you can add structured, typed metadata (like priority, effort, dates, and custom values) to issues across your repos."

"Blume reads your agent conversations locally, then promotes steering and pain signals into code to make the next run better and cheaper. Stop repeating yourself to coding agents"

"SQLAdmin provides an admin interface for SQLAlchemy models that works with Starlette and FastAPI. Sync/async SQLAlchemy engine support, WTForms-based form building, SQLModel compatibility"

"nodenv automatically selects the correct Node.js version for each project by scanning for a .node-version file, with no performance overhead."

"The mega bull case for AI infrastructure would be *if* market share shifted away from certain frontier labs with 90%+ inference margins toward cheaper models... Lower margin % at the model layer = more margin $ at the infra layer"