Rewrite LLM-drafted essays to sound like Steve Krouse using steve-eval.md as a scoring rubric.

1. Score the current draft against steve-eval.md. Produce a detailed scorecard with quotes for every sub-criterion and a total. Identify the biggest gaps.
2. Rewrite the essay targeting the lowest-scoring categories first. Refer to the rubric constantly.
3. Re-score the rewrite. Produce another full scorecard.
4. Iterate until the score is 95+.

Codex "Score intelligence.md against steve-eval.md, rewrite it to score higher, and iterate until 95+."

Or more generally:

Codex "Score <essay>.md against steve-eval.md, rewrite it to score higher, and iterate until 95+."

• Always commit before returning to Steve. Ensure all changes are committed before starting new work.
• Do real work in subagents so the main conversation is free for discussion and management.

• steve-eval.md – Voice rubric (0-100). 90+ = almost certainly Steve.
• <essay>.md – The essay being rewritten.

• If you want to know how to edit vals on Val Town, here's their agent instructions: https://www.val.town/townie/system-prompt
• The HN mod sent us advice for writing front-page HN posts: see references/hn-frontpage-advice.md (source: https://gist.github.com/stevekrouse/22d96b6142fd21a8f9b63500ebd2ee24).