Last week
Agent costs demand hard budget caps, not warningsOct 5Week of September 28
Claude Code adds mods, including a self-monitoring agentOct 2Claude Code can now auto-build your evalsOct 1Opus 5.5's prompting guide breaks some old patternsSep 28Week of September 21
Benchmark scores often aren't reproducible, UK AISI findsSep 25Caching becomes an explicit, controllable API primitiveSep 24Agentic iteration made Rust code faster than expertsSep 22Passing once doesn't prove an agent is reliableSep 21Week of September 14
Mollick: pair deep expertise with AI's breadth for an edgeSep 20Ptacek: use an LLM as copyeditor, never as ghostwriterSep 18Messages API can now compact a conversation on demandSep 17AWS: 'production cage' = scoped capabilities + blast-radius cap per agent taskSep 15Vercel: agent building is a product problem, not a model problemSep 15MCP servers: expose workflows, not one tool per endpointSep 14Week of September 7
Claude Code names containment escape; effort gets a capSep 13Managed Agents now auto-evaluates each tool call individuallySep 11Fable 5.1 thinking blocks persist across turns — and bind to your prefixSep 11Fable 5.1 cache reads dropped 75% — now $0.25/MTokSep 10`clear_at` lets per-turn reminders render once and cost nothing afterSep 10`ant apply` ships — Claude agents and skills deployable from YAMLSep 8