Build plan · v1 · Review & decide

NLP Mastery — Claude vs Codex duel build.

For each task, tap YES to approve, NO to skip, or MODIFY and write what you actually want. Decisions auto-save back to the server. When you're done, hit Submit at the bottom — Claude will then build everything you approved and report back what's outstanding.

Builders Claude + Codex, same brief Lab /opt/nlp_duel Based on edge_crypto pairing

Yes

Modify

Pending

0 reviewed

How each milestone runs

Build: both build in parallel, separate sandboxes, neither sees the other.
Score: frozen visible tests + hidden judging tests.
Cross-review: each audits the other and writes what's wrong and what's better.
Improve: each fixes its own build and may borrow ideas, then re-score.
Report: saved to /opt/nlp_duel/reports/; TLDR when you ask.

Rough timing

Setup + brief + tests: about 1–1.5 days
Each milestone: a few hours of agent time, running unattended
Main wait is your brief approval and any questions the builders raise

Submit final decisions?

Once submitted, Claude reads your responses and starts executing. You can still come back and edit, but rerunning the build manually after edits is on you.