Expanso team: no more code review, tests or CI — just writing evals
aronchick · x · 2026-10-03
An Expanso engineer describes the team's internal reality with AI: they discussed how they barely review code anymore — no tests, no CI, almost nothing to write — leaving only eval-writing. Quoted reply adds that capability inflation has "completely cooked" calibration: things that felt like witchcraft 18 months ago now infuriate when they take 20 minutes. A snapshot of how high-skill teams shift verification from code review toward evals in agentic coding workflows.
More from coding & agent
- jasonkneen: every personal agent has the same unsolved problem — jasonkneen · 2026-10-03
- Dev builds WoW private server, browser client, and an MCP agent to play it — ProfessorMunchy · 2026-10-03
- Y Combinator's QM team-oriented AI agent harness is now free on Agent37 — ycombinator · 2026-10-03
- OpenAI's Dot Agent Spotted Production Outage Five Minutes Before DevDay Live Demo — lennysan · 2026-10-03
- LangChain's Chase Amps Up 'Durable Harness' Debate as Pi Durable Draws HN Praise — hwchase17 · 2026-10-03
- OpenAI Dot's Codex task access only works about 50% of the time, user reports — Danny7Guns · 2026-10-03