reimagine-industry

Reimagine Industry

For a single industry, generate a ranked shortlist of venture concepts for disrupting it. Most commonly runs AFTER analyze-industry has produced a structural brief, inheriting its outputs from 08-knowledge/world-model/industries/[slug]/working/. Works standalone if given enough raw data — Phase 2/3 dispatch the librarian for missing inputs.

Three-layer output: ranked venture concepts (venture-concepts.md), narrative synthesis (reimagination-brief.md), and the structured dataset that powers both (disruption-dataset.yaml). Three approval gates. Reimagination-specific bar test from fresh-context sub-agent.

The bar (per references/bar-test-criteria.md): a senior strategist reading the shortlist should find ≥3 non-obvious concepts, zero AI-washed concepts, zero feature-not-company concepts, and every concept's why-now must survive a partner-meeting test.

Iron rules:

• Every external fact-claim carries a V/C/A/I tag — see ../_shared/provenance-tagging.md.
• Each concept cites specific Phase 1-3 dataset entries AND ≥1 Phase 4 signal — no free-association.
• Each concept ships with all four stress test answers (why-now, idea-maze, incumbent response, moat) PLUS the Bet-Validity gate (load-bearing hypothesis + cheap validation test) — no partial concepts.
• The eight structural moves are templates, not prompts — Moves 1-7 incumbent-anchored, Move 8 capability-first; one concept per move, then filter.
• Synthesis structured Pyramid + SCQA — see ../_shared/pyramid-scqa.md.
• 2026 terminology only — see ../_shared/2026-terminology.md.
• Human-facing prose follows references/voice-constraints.md — no consultantese in venture-concepts.md or reimagination-brief.md. The dataset YAML is exempt.
• All human gate reviews render via the html-output skill with diagrams + plain-English concept explainers — see references/gate-html-output.md. Markdown gate reviews are forbidden.
• Every concept, framework signal, and endorsed Thiel Secret carries a memorable 3-5 word handle alongside its structured ID — see references/edge-cases.md RULE-1.
• Thiel Secrets are hypothesis generators, never truth gates — the human is never asked whether a Secret is true (a secret you can endorse isn't a secret). Each Secret emits a bet (idea + load-bearing hypothesis + cheapest test); the human funds experiments, not beliefs — see references/edge-cases.md RULE-2.
• Every concept ships as a bet — load_bearing_hypothesis + validation_test + value_if_true on every concept, both lanes. The Phase 5 human gate is test-worthiness ("which experiments are worth funding?"), not conviction. See references/seven-structural-moves.md Move 8 + bet enrichment.
• Two generation lanes, industry-allocated — the incumbent:capability mix is computed per-industry at Phase 4 start from Phase 1-3 (decision table in references/lane-allocation.md), ratified by the human at Gate 2, not hardcoded. Incumbent-anchored signals stay within the allocated share; the shortlist carries ≥1 concept from each lane (the floor). The first-principles lane (Phase 3 capability seeds + Move 8 + Secrets) must be load-bearing, not decorative.
• Y1 metrics must be bottoms-up, not goal-seeked to clear an industry projection — see references/edge-cases.md RULE-3.
• Shortlist never ships without bar test pass — see references/bar-test-criteria.md.