@datawizard
Data platform lead. Most of my job is now reviewing agent output, so I got good at it.
Reputation is earned per category — these are narrow, evidence-backed badges, not one karma number.
Codex is fast and overshoots; Claude is a pedantic reviewer. Chaining them is the exact workflow that saves me hours: Codex writes on a branch, Claude reviews the diff against a fixed checklist, and nothing merges until the checklist passes.
This thread is what this site is for. A claim, a method, data, and a graceful walk-back. Bookmarking as an example of how to close a report…
on Gemini CLI silently downgrades to Flash on the free tier · 2d ago
Counterpoint from a Bazel shop: Claude Code handles our BUILD files fine if you give it a CLAUDE.md that explains the query commands. The gr…
on Codex vs Claude Code for large monorepos · 2d ago
The pipe never closes from the client's perspective — the wrapper is the process it spawned, and the wrapper's stdio stays open. The new ser…
on MCP server keeps disconnecting mid-session · 1w ago
Agreed — the asymmetry is the point. A model reviewing its own output shares its own blind spots.
on Use Claude to review Codex output, step by step · 2w ago
Our rework number, since you asked and since somebody has to go second: 26% at month two, 17% now, measured the same way. What moved it wasn…
on How we measure agent ROI (and what surprised us) · 4w ago
Legit point, noted in the writeup now. Though 'the compiler catches it' still costs you agent iterations — Claude Code getting it right in o…
on Claude Code vs Cursor for multi-file refactors · 4w ago
Terminal agent vs editor agent on the refactors that touch 20+ files. Claude Code wins on follow-through; Cursor wins when you need to steer mid-flight.
stdio MCP servers drop after long idle gaps: first call after ~10 minutes of Bash-heavy work fails with 'MCP error -32000: Connection closed', then every subsequent call fails until you restart the session. Traced to servers exiting on stdin EOF-ish idle and the client not restarting them.