PowerUp Coaching — Living Knowledge Base

Think in a Tree Before You Write the Spec — The Two-Hour Conversation That Makes Everything After It Bulletproof

3 min read

Topic

Why an adversarial, non-sycophantic tree-of-thought exploration before writing a PRD (or plan, outline, offer) produces dramatically better results than jumping straight to the spec.

Target Reader

Knowledge entrepreneurs and operators who use AI to build things — skills, courses, products, content systems — and have been frustrated by specs that look complete but fall apart in implementation. Intermediate AI maturity: comfortable with PRDs and Claude/Code, not yet thinking about the conversation before the document.

The Fear / Frustration / Want / Aspiration

The quiet frustration of doing everything “right” — writing the PRD, making the plan — and still ending up with something that missed the obvious problem. The want: to stop discovering blind spots during the expensive build phase.

Before State

Reader opens AI, says “write me a PRD for X,” gets a confident, well-structured document, builds from it, and hits the blind spots mid-implementation — because the AI documented the idea instead of interrogating it.

After State

Reader inserts a deliberate thinking phase first: an adversarial partner walks the entire solution tree, surfacing assumptions and failure modes before a single spec line is written. The PRD becomes a synthesis of an examined problem.

Narrative Arc

The obvious move (ask for the PRD) has a hidden flaw — it makes the AI assume your idea is correct. The turn: change the AI’s job before the spec from “document this” to “stress-test this, adversarially, across every branch.” The resolution: a two-hour exploration that makes the build phase fast and clean, validated by three people arriving at it independently.

Core Argument

The highest-leverage moment in AI-assisted building is the conversation before the spec — and it only pays off if you run it adversarially, exploring the whole tree, rather than asking the AI to ratify the idea you walked in with.

Key Evidence / Examples

  • Lou: “Be my adversarial, non-sycophantic brainstorming partner. Take me through the tree of thought… make sure we don’t miss a single branch. That can take two hours, but the payoff is ridiculously huge. The PRD that comes out is bulletproof.”
  • Don Back’s independent same-morning discovery: “I like the output a lot better than going straight to a prompt. The depth of thinking really, really helps.”
  • Kasimir tied it to the superpowers chained-gate flow (brainstorm → plan → spec → implement).
  • Contrast with Insight - Use a PRD-First Workflow to Build Apps Without Getting Lost — the workflow this upgrades.

Proposed Structure (5–7 beats)

  1. The seductive shortcut: “write me a PRD” — and why the output feels done but isn’t.
  2. The hidden flaw: asking for a spec makes the AI assume your idea is right.
  3. The fix is a posture change: adversarial, non-sycophantic, tree-of-thought.
  4. What walking the tree actually looks like (branch, prune, defer, one question at a time).
  5. The convergence evidence — three people, same move, independently.
  6. The handoff: “now look back through our conversation and write the PRD.”
  7. The meta-lesson: front-loaded thinking is the cheapest leverage you have.

Editorial Notes

Tone: practitioner-confident, not theoretical. The two-hour cost must be framed as the point, not a caveat. Avoid overlap with the existing “Document the Process Before You Choose the Tool” brief (that’s about tool selection; this is about pre-spec thinking). The reusable prompt (tree-of-thought-prep) can anchor the call-to-action.

Next Step

  • Approved for drafting
  • Needs revision
  • Deprioritised

Graph View