1. July 2026
  2. An AI agent reviewed his release and found a silent data-loss bug (simonwillison.net)
    AI · · July 7, 2026 · Agents · Engineering · Claude
  3. Claude has an internal workspace, and Anthropic built a tool to read it (anthropic.com)
    AI · · July 7, 2026 · Anthropic · Interpretability · Research · Safety
  4. Why price per million tokens tells you almost nothing (janilowski.pl)
    AI · · July 7, 2026 · Economics · Evaluation · Models
  5. AI out-persuades expert humans, but the edge is speed, not eloquence (importai.substack.com)
    AI · · July 6, 2026 · Safety · Research · Persuasion
  6. Newer Claude models are worse at custom tool schemas (simonwillison.net)
    AI · · July 6, 2026 · Anthropic · Agents · Tool Use
  7. Claude Code's guide to loops: the hard part is defining done (claude.com)
    AI · · July 6, 2026 · Claude · Agents · Engineering
  8. Mistral's Leanstral 1.5 saturates a math-proof benchmark and finds real bugs (mistral.ai)
    AI · · July 6, 2026 · Mistral · Open Models · Formal Verification
  9. How long until AI can run without humans? Cotra and Lee disagree by decades (asteriskmag.com)
    AI · · July 6, 2026 · Forecasting · Robotics · Economics
  10. GeneBench-Pro shows AI still stumbles on real biology analysis (openai.com)
    AI · · July 5, 2026 · OpenAI · Evaluation · Science
  11. Zuckerberg tells staff Meta's AI agents are behind schedule (techcrunch.com)
    AI · · July 5, 2026 · Meta · Agents · Economics