Skip to content
DecisionNerd Skills

Discover

Learn before you pour accelerant. Discovery is not a phase you finish once — it re-enters from Operate feedback and Maintain bets.

Continuous discovery

Section titled “Continuous discovery”
  • Frame the problem / opportunity before solution theater.
  • Gather evidence from users — and for DX, treat developers and agents as users of the system you build.
  • Loop: observe → learn → decide the next bet. Tie to observe-it, ProductFeeling, and DocSlime experience/ when you earn that altitude.

Do not confuse “we brainstormed” with discovery. Discovery produces falsifiable next bets.

Fidelity: pretotype → prototype → MVP

Section titled “Fidelity: pretotype → prototype → MVP”

“MVP” means three contradictory things in the wild. Name the level you’re actually at:

Level Question it answers Typical artifact
Pretotype Would anyone care / try? Fake door, landing + waitlist, paper, Wizard of Oz
Prototype Can this interaction / tech approach work? Clickable or throwaway code; learning > longevity
MVP — utility-proving Does this create real value even if ugly? Thinnest real usage path
MVP — usable Can a real user complete the job without heroics? Minimum usable product
MVP — shippable Can we operate and change it without lighting new fires? CI, ownership, recoverable deploy

House stance: pick the lowest fidelity that can falsify the risk you’re actually carrying. Run kiss when fidelity is unearned. Selling shippable while delivering a pretotype is framing debt — see Bugs & debt.

Spec altitude: rough vs DocSlime

Section titled “Spec altitude: rough vs DocSlime”

Two honest modes (not a moral hierarchy):

  • Exploratory building — rough specification: vision sentence, 2–6 BDD completion scenarios, done-when, known unknowns. Prefer surgical document-it. Do not scaffold full DocSlime theater.
  • Contract building — fuller DocSlime altitude when stakes, multi-person, regulated, or agent-heavy work needs a durable quality trace. Use docslime-* when earned.

Coffee test: What failure is expensive if we’re wrong — interest, trust, or infra? Higher cost → higher altitude.

Agents still need falsifiable scenarios in rough mode; full DocSlime helps them share an oracle — same quality idea, different altitude.

Next

Section titled “Next”