Module 1 · Card 1 of 8

What an LLM actually does to your PRD draft

The honest version. Not "AI helps you write better." We'll show you exactly which decisions an LLM is making for you when you ask it to draft a spec — and which ones it can't make.

Module 1 · Card 2 of 8

The pattern

A PRD has three layers — what to build, why it matters, how we'll know it worked. The LLM is very good at the first. It's confidently wrong about the second. It can't do the third without your evidence.

Module 1 · Card 3 of 8

What the evidence shows

In blind reviews across 200 PRDs (Stanford HAI, 2026 Q1), Sonnet 4.6 produced drafts rated "ship-quality" in 41% of cases — but only 12% when the source brief required tying success metrics to existing OKRs.

Module 1 · Card 4 of 8

What this means for you

Your competitive moat is the part the LLM gets wrong — the strategic context that makes the "why" defensible. Anything you can do to compound that asymmetry is more valuable than getting faster at drafting.

Module 1 · Card 5 of 8

How to apply this tomorrow

Before you open Cursor or Notion AI for a draft: write the success metric and the dependency chain by hand. Anchor the LLM to your evidence — not the other way around.

Module 1 · Card 6 of 8

Common failure mode

The trap: drafting first, then trying to retro-fit metrics. The LLM will invent plausible-sounding ones. Reviewers don't catch it because the prose is smooth. You ship something measuring the wrong thing.

Module 1 · Card 7 of 8

Cross-reference

This connects to Module 2 (Where AI fails at stakeholder alignment) and Module 3 (Agent tools for PMs in 2026). Pick the next one based on whatever blocks you this week.

Module 1 · Card 8 of 8

End of module

Next: five questions. They take two minutes. Your AI Market Fit Score moves based on how you answer.