1,500 real coding-agent sessions measured: 17% of spend wasted on retries
entelligenceai17 · reddit · 2026-10-27
Entelligence analyzed their own Claude Code and Codex session logs with AgentPlay, an open-source local log-analysis tool. Key findings: only 80 of 912 code-editing sessions ran a check after the last edit; 17% of spend went to re-reads, retries and abandoned work; one MCP server failed 84% of its calls yet was still called 117 times. Runnable via uvx agentplay@latest.
More from coding & agent
- Chat Brawlers: three AI models debate and vote — and one tricky stopping rule — lluisantoni · 2026-10-28
- Dumb chunk-and-rerank baseline beats top agentic memory systems in 6 of 7 benchmarks — JeremyCMorgan · 2026-10-28
- Microsoft's MANDATE: 23 failure categories from 163k Reddit posts on AI agents — dair_ai · 2026-10-28
- LangSmith now tracks cache hit rates for your agents — LangChain · 2026-10-28
- MOSS-Image ecosystem roundup: ComfyUI 0.43, QwenVL layout node, new style LoRAs — optimisticalish · 2026-10-28
- Ex-PC repairman builds JinnexGuard with Claude Code to scan MCP servers before install — Complex-Pin-4616 · 2026-10-28