StartupENGINEERING
Loop 5 of 5 · The Five Loops

The Fit Loop

North star: “The business repeats without founder heroics.

At hand
Retention

The product plus its returning customers — retention, being read.

What you hold is the tool of discovery — every loop's input, never its output. The name tells you what you're hunting; your hands tell you where you're standing.

North star

The business repeats without founder heroics.

Entering assumption

The market validated — customers arrived; unknown whether they return without you.

Engineering bar
Repeatable System

A repeatable system — the product plus its operations, minus your heroics. What repeats at 1 isn't the software; it's the system around it.

  1. 1 · Problem LoopManual Delivery
  2. 2 · Solution LoopDisposable Prototype
  3. 3 · Product LoopLaunchable MVP
  4. 4 · Market LoopObservable Product
  5. 5 · Fit LoopRepeatable System
The engineering bar rises only when the evidence earns it.Beyond 1: Scalable Platform — off this ladder, by design.
Exit bar
Observable evidence to leave. None of it is a feeling.
  1. 1The same customers return, and the value moment fires again — without founder heroics.
  2. 2The retention curve flattens instead of decaying toward zero — the ones who stay, stay.
  3. 3It repeats at a serve cost you have named, covered by the price they pay.

1 is the day the business repeats without you. This is the evidence on which founders commonly declare product-market fit — the declaration is always the market's to keep confirming; what you hold is the repetition.

Earns: 1 — the business repeats. All five exit bars →

First principles
  • Repeatability is the product of subtraction.

    Every delivery step that is currently you, made into product or process, in order of frequency. Automation is not a technology project but a transfer of heroics.

  • Technical debt gets a triage rule.

    Debt that blocks repeatability is paid now; debt that merely offends engineers waits. The ledger from the Product loop finally comes due — selectively.

  • Unit economics gain their honest final row.

    Infrastructure cost plus founder-hours, priced, per customer, against revenue. A margin that only exists at zero founder-hours you don't have is an opinion, not a business.

This loop in the AI era
Revision surface — rewritten when the wave changes. The framework beneath it doesn't move.
What changed

AI is this loop's best subtraction engine — the repeating exceptions, reports and follow-ups that were your heroics are what today's tools automate first, each started at the “drafts” notch and promoted on evidence. And per-action AI cost becomes a unit-economics question with teeth: the business is viable only if the model bill per customer survives your price.

What didn't change

Repeat is still proven by customers, not capabilities — no model release renews a contract.

Embrace

Automating your own heroics before hiring for them.

Ignore

The temptation to call AI leverage “traction” — the vacation test doesn't care what did the work, only whether you had to.

Decisions in this loop
Browse Fit Loop decisions in the catalogue →
Not Yet

Decisions explicitly deferred while you're in this loop. Not wrong questions — early ones. Written down, and set down.

  • Scaling architecture ahead of load
  • Team building beyond the first hire
  • Everything beyond 1 — growth and scale-up get their own engineering

When Not Yet becomes Now: an obligation crosses over only by naming its class and pointing at its evidence — contract-earned, load-earned, risk-earned, or repeatability-earned. Anything else is still anxiety-driven, and stays Not Yet. The sixth rung is climbed by demand, never by dread.

Stamp your own worries with the Loop Check →

Decision skills

Free, ungated plain-text files you hand to Claude or ChatGPT. All skills →