Pick a depth. Each prompt opens in your AI pre-loaded with the lesson. Click a row to preview the prompt.
Perfectionism is the slow sin. Sin #8 (new-feature-itis) and Sin #9 (polishing past the point of value) together produce projects that never ship — or ship 6 months late with features no one asked for. The cure is brutal scoping discipline: a small, named set of features in the v1 box, and an explicit cutoff for polishing. Cohen's insight: every additional feature past the core MVP costs disproportionately more time than it adds value.
The scope creep ratchet and the polishing curve.
SIN #8: NEW-FEATURE-ITIS
The pattern:
v1 plan: features A, B, C. Ship by date X.
Month 3: "We should add D — it's only a small change."
Month 5: "D is great. Now let's add E for parity with competitor."
Month 7: "While we're at it, F would be trivial..."
Month 9: Original date X has passed. Features D-F not done.
Why it happens:
- Adding feels productive (more is more, right?).
- Sales/marketing keep adding requests.
- Engineers want to build the cool thing they imagined.
- PM can't say no.
Cohen's cure:
- A locked v1 feature list with names attached.
- A change control board: any addition requires explicit acceptance of a slip.
- "We can do D in v2" should be the default answer.
SIN #9: NOT KNOWING WHEN TO QUIT POLISHING
The pattern:
Product 90% done at month 5.
Months 6, 7, 8 spent on:
- Cosmetic UI tweaks
- Edge cases that affect <1% of users
- Performance optimization no one asked for
- "Just one more iteration of the case design"
Why it happens:
- Fear of shipping (and being judged).
- Pride: "I want it to be perfect."
- Lack of pressure from a clear ship date.
Cohen's cure:
- 95% rule: ship at 95% quality, not 100%.
- Time-box polish: 2 weeks max after v1 feature-complete.
- User testing > polish: real users find bugs you'd never have polished anyway.
HARDWARE TRACK:
Cohen's example: connected device team adds 3 features in months 3-5
(scope creep), then spends months 6-8 polishing the enclosure shape.
Ships at month 9. v1 had launched, market reception was lukewarm,
competitor shipped at month 7 with worse hardware but identical UX.
The hardware penalty for polishing is large:
- Each enclosure revision: $10-30k tooling spend.
- Each PCB revision: 4-8 weeks + $1-3k.
- Polish in v1 → no money for v2 improvements.
SOFTWARE ANALOGUE:
Software has the same sins but lower revision cost. The pattern shifts to:
- Scope creep: backlog grows infinitely; sprint scope keeps swelling.
- Polish: "we'll redo the design system before launch" syndrome.
Software polish trap:
- 5 designers in Figma at month 6, no users in production yet.
- 3 refactor passes on the same code path before any user touches it.
- "Performance optimization" before there's any load.
Better:
- Ship at "good enough." Iterate on what real users complain about.
- Linear-rule: backlog is a graveyard; only what's in the current
cycle counts.
THIS MODULE'S FOCUS:
Two questions to ask yourself this week:
1. What was in the v1 list 90 days ago? What's in it now? (Sin #8)
2. If we shipped today at 95%, what would actually be missing? (Sin #9)
The answers usually reveal one of the two sins live.