Schema as specification: making the 70/30 split visible

Frontmatter

{
  "id": "petrova-wave3-schema-as-specification",
  "type": "design-journal",
  "stage": "drafting",
  "title": "Schema as specification: making the 70/30 split visible",
  "author": "yao",
  "claims": [
    {
      "text": "teams winning at agentic coding spend 70% on problem definition, 30% on execution",
      "source_ref": "70_30_split_to_petrova-1"
    }
  ],
  "created": "2026-06-07",
  "arc_thread": "petrova-addy-cognitive-harness-arc",
  "created_at": "2026-06-07T00:00:00Z",
  "what_changed": "The claim-gate now verifies cited claims against a pinned claim index before content can advance.",
  "last_modified": "2026-06-07T00:00:00Z",
  "target_domain": "devarno.com",
  "what_was_kept": "The resolver stays structural-only and fail-safe; semantics live in the verifier.",
  "primary_system": "petrova",
  "why_it_changed": "Parseable-but-unbacked claims could previously pass; the gate makes verification the default.",
  "decision_gating": {
    "campaign-2026-05-08-stratt-integration": "light-mention (pending Devarno approval — swap if decided otherwise)"
  },
  "systems_affected": [
    "petrova"
  ],
  "decision_recorded": "PETROVA treats the schema as the problem definition; everything after validation is execution.",
  "minimum_dwell_met": false,
  "word_count_target": 1000,
  "lexicon_terms_used": [],
  "transition_history": [
    {
      "actor": "yao",
      "reason": "staged from campaign-content-templates (Wave 3) during sub-project A",
      "to_stage": "drafting",
      "from_stage": null,
      "transitioned_at": "2026-06-07T00:00:00Z"
    }
  ],
  "vulnerability_check": "Fail-safe: a missing/invalid index yields all-unverified, not silent pass.",
  "anti_patterns_checked": [],
  "what_might_change_next": "Retroactive re-verification of queued artifacts when the index changes (deferred)."
}

Body

## Design journal — schema as specification

Addy measured it: teams winning at agentic coding spend 70% on problem
definition, 30% on execution. PETROVA makes that visible. The schema *is* the
problem definition. Everything after validation is execution. When you spec the
schema correctly, the agent can't misinterpret the problem.

That's not a limitation. That's architecture.

(STRATT gets a light mention in the audit-trail section, pending the
stratt-integration decision.)