Open-source Rust tool ai-memory (MIT, ~7.6k stars) stores agent memory as a git-backed markdown wiki over MCP/HTTP, so you can leave Claude Code mid-task and continue in Codex or Cursor without re-briefing. The default write path needs zero LLM calls; 20+ harnesses are supported.
White hat firm Hacktron chained a Discourse/libheif image bug with an OpenAI SSO flaw, then used an employee’s Codex session to open a harmless PR in an internal repo—under 72 hours end to end. Claude Opus 5 built the working exploit overnight after Opus 4.8 failed; OpenAI patched in 14 hours and paid a $6,500 bounty…
British Columbia filed a U.S. federal suit against OpenAI and Sam Altman on Monday, alleging the company flagged the Tumbler Ridge shooter’s ChatGPT chats for gun violence but did not alert RCMP before the February attack that killed eight. Altman apologized in April for not notifying law enforcement; families have separate suits underway…
An independent Advisory Group on Mathematics and AI, hosted at IAS and announced on Terence Tao’s blog, will advise labs on releasing math results without pay from companies. Its first job is helping OpenAI coordinate disclosure of many significant results an internal model reportedly produced…
Tim Dettmers’ dlab Open Source Week teases local inference (Qwen 3.6 35B A3B at 450 tok/s on Mac at 1.5 bpw; 125B on one 24GB GPU; DeepSeek V4.1 550B on 128GB boxes) plus CliffCompaction that cuts agent cost 50% vs Claude Code/Codex style compaction. Two open source projects and four papers drop as one ecosystem, delayed one day from the Sept 21 post…
Maharshi Patel’s essay arguing MCP was built for weaker 2024 models and should yield to shell plus HTTP APIs hit Hacker News with 320 points. Critics including Simon Willison say MCP still matters for auth boundaries, audit logs, and agents without unrestricted shell access…
Spotted something we missed? Start a thread.
