“Before I wrote the PRD, I realized the purpose of that conversation was just to create the PRD — it assumed the thing I’m asking for was actually required. So now I take a step back: ‘Be my adversarial, non-sycophantic brainstorming partner. Take me through the tree of thought. Explore every branch — every feature, every problem, every blind spot — and make sure we don’t miss a single one.’ That can take two hours. But the payoff is ridiculously huge. The PRD that comes out is bulletproof.” — Lou
Session context: 2026-06-18_Mastermind — Lou walked through the process behind his 15-hour agentic-writing-team build; Don Back independently reported discovering the same move that same morning.
Core Idea
The PRD-first workflow has a hidden flaw: if you jump straight to “write me a PRD,” the AI’s job is to document the thing you asked for — not to interrogate whether it should exist. It treats your idea as a given and writes the spec for it, blind spots and all. The fix is to insert an explicit thinking phase before the PRD, with a different posture and a different goal.
The posture is adversarial and non-sycophantic: “push back on anything I say that doesn’t make sense, and don’t cave on anything you believe in.” The structure is tree-of-thought: explore every branch of the solution surface — features, failure modes, blind spots, technical requirements — and open a new branch whenever something surfaces, until the whole tree is walked. Only then do you say “look through our conversation and create a PRD.” The PRD is now a synthesis of a fully-explored problem, not a transcription of an unexamined idea.
The strongest evidence is convergence: two members arrived at the same move independently. Don Back ran the exact pattern on a writing skill that morning — exploring and debating before moving to PRD, then to skill — and said: “I like the output a lot better than what I would have done in the past, which would be going straight to a prompt. The depth of thinking really, really helps.” Kasimir tied it to the superpowers skill’s chained gates (brainstorm → plan → spec → implement, each confirmed before the next). When three people converge on “think harder up front, in an adversarial frame, before you spec,” the practice has transcended personal style.
Practical Application
Next time you’re tempted to ask AI to “write the PRD” (or the plan, the outline, the offer), don’t. First run this setup prompt:
“I have an idea. Be my adversarial, non-sycophantic stress-test partner. Take me through a tree of thought — explore every branch, every feature, every problem, every blind spot. Push back on anything that doesn’t hold up; don’t cave on anything you’re sure of. Open a new branch any time something surfaces. Don’t let us miss a single one.”
Walk the tree to exhaustion. Then say: “Now look back through our whole conversation and write the PRD.” Defer dead-end branches to a separate chat with a one-line note, so curiosity doesn’t derail the main thread.
Related Insights
- Insight - Use a PRD-First Workflow to Build Apps Without Getting Lost — the workflow this upgrades; the new move is a thinking phase before the PRD.
- Insight - Tell It What to Do, Don’t Ask It Questions — The Posture That Makes AI Think — the adversarial, directive posture that makes the exploration productive instead of agreeable.
- Insight - Spec-First Vibe Coding - Multi-Model Architecture Review Before Writing a Line of Code — the same “think before you build” principle at the architecture level.
- Insight - Rewind, Don’t Re-Correct — Keep Failed Attempts Out of the Context Window — the context-hygiene technique that keeps a two-hour exploration high-signal.
Evolution Across Sessions
Extends Insight - Use a PRD-First Workflow to Build Apps Without Getting Lost (2025-09-11). That insight established: document before you build. This one inserts a prior stage — adversarially explore the entire solution tree before you document — and names why the obvious shortcut fails (asking for a PRD makes the AI assume your idea is correct). First articulation with multi-member convergence, which marks it as a mature, proven practice rather than one person’s habit.