red-team

# Red Team Adversarial review: steelman the plan in its strongest form, then attack it like a skeptical exec, a rival, and reality itself would. The goal is to surface the assumptions and weak links that decide success — while it's still cheap to fix them. **Grounded in:** *Red Teaming* — Bryce Hoffman: attack the plan's load-bearing assumptions and diagnosis before reality does. **Go deeper (The Product Channel):** [Why Every PM Needs to Master AI Evals](https://sidsaladi.substack.com/p/why-every-pm-needs-to-master-ai-evals) ## When to use this - You're about to send a PRD, strategy memo, or business case to leadership and want it bulletproof before someone else finds the holes. - A plan "feels right" and nobody's pushed back — which usually means nobody has looked hard enough. - You're prepping for a tough review (exec, board, investor, architecture) and need to know the questions before they're asked. - A bet is large, hard to reverse, or rests on a few key assumptions you haven't tested. - A competitor or regulator could undercut the plan and you want to game out their move first. ## Before you start (gather these) - **The artifact itself** — the full PRD, strategy doc, or plan text (not a summary). - **The core claim** — what this plan asserts will happen and why (the thesis in one sentence). - **Stakes & reversibility** — how big the bet is, how hard it is to undo, who's accountable. - **The skeptic** — who will attack this in real life (CFO, eng lead, competitor, c...