Write Section Promise

Four tests

A promise passes if ALL of:

1. Specific — reader can predict what a post in this section will do for them.
2. Testable — point at a candidate post and say yes/no.
3. Non-overlapping — distinguishable from every other section's promise by at least one axis.
4. Voiced — reads as the writer's voice per audience-notes.md + voice-profile.md, not marketing copy.

Workflow

For a candidate section:
- [ ] Step 1: Draft 3 candidate promises
- [ ] Step 2: Score each on the 4 tests
- [ ] Step 3: Pick highest-scoring
- [ ] Step 4: If none pass all 4, flag and return best-with-caveat

Worked example

Section: Intuition-first ML (candidate).

• Draft A: "Essays about machine learning." — FAILS specific, FAILS non-overlap (too broad).
• Draft B: "We break down ML concepts." — FAILS voiced (marketing tone), FAILS testable.
• Draft C: "Essays that build geometric or mechanical intuition for how ML systems actually work, before the math." — PASSES all 4. Winner.

Guardrails

1. ≤25 words. Longer promises become paragraphs and no one reads.
2. No buzzwords. No "we dive deep into X" or "unlock insights."
3. One sentence, not two.
4. Second person or declarative — never "I argue…"
5. Must fit the writer's voice per audience-notes.md + voice-profile.md.