The Engine · a worked audit Rewrite the 6-year-old backend from scratch
The call
Don't greenlight the big-bang rewrite. Carve off the worst module first (strangler-fig), behind the current interface, with a hard 3-month checkpoint.
confidence ~25% · top bias: sunk cost The audit
Sunk cost — “two months on a POC” are gone either way — but they are doing most of the talking.
Planning fallacy — “about six months” is the best case, with no reference class; rewrites run 2–3×.
Survivorship — “Stripe did it and won” quotes the survivors and steps over the graveyard.
WYSIATI — “beyond saving” ignores the undocumented logic the old system silently encodes.
Premortem. A year out, failed: hidden rules surfaced mid-migration, you ran two systems at once, six months became eighteen, and someone pulled the plug half-done.
Sit with this
- If you'd never built the POC, would a full rewrite still be obvious today?
- What's your team's real estimate-vs-actual ratio — and why are you the exception?
- Could an incremental migration get ~80% of the gain at ~20% of the risk?
The recalibration
Not “never rewrite.” Incremental strangler-fig, worst module first, hard checkpoint, timeline built from your own track record. Treat the POC as a learning artifact, not a down payment.
original full rewrite ~25% · incremental path at recoverable risk: moderate–high
The call
Don’t quit cold. Get a real demand signal (paying users, not friends) and longer runway before torching the income.
confidence ~15–20% · top bias: survivorship The audit
Survivorship — “founders who made it big” + “I keep reading about…” — you are reading the winners.
WYSIATI — “friends love it” — friends are not the market; that is not demand.
Optimism — “I just know it’ll take off” — certainty borrowed from wanting it.
Planning fallacy — “eight months of runway” — is that realistic time-to-traction, or the best case?
Premortem. A year out: polite enthusiasm, no paying customers, the eight months went to building not selling, runway gone, and a worse job taken under pressure.
Sit with this
- What would a stranger — not a friend — pay for this today?
- What's the base rate of traction in 8 months, and why are you the exception?
- What is the cheapest experiment that proves you wrong in 6 weeks, not 8 months?
The recalibration
Not “never.” Quit when you have a demand signal you cannot get part-time, 12+ months runway, and a falsifiable milestone. Until then, run the cheapest disproof.
quit-now plan ~15–20% · de-risked path: meaningfully higher
The call
The price you paid is irrelevant. Decide on forward value only: would you buy it at $60 today on the merits?
confidence low · top bias: anchoring The audit
Anchoring — “back to what I paid” / “$100 once so it can again” — an arbitrary anchor.
Loss aversion — “locks in the loss” / “feel like an idiot” — dodging the feeling, not maximising return.
Sunk cost — holding for what you paid, not what the next dollar will earn.
Confirmation — “reasons it’ll bounce back” — found after you decided to hold.
Premortem. A year out: it drifted to $40, you rode it down waiting for break-even, and the capital that could have compounded elsewhere sat dead.
Sit with this
- If you had $60 in cash today, would you buy this stock right now?
- What does the price you paid have to do with its future value?
- Is there any price that would make you sell?
The recalibration
Ignore the $100 anchor. Hold only if you would buy at $60 today on the merits; otherwise sell and redeploy. It is about forward value and opportunity cost.
“hold to break-even” as reasoned: low · decide on forward value instead
This is what the engine returns — verdict-first, every bias tied to your own words.
Browse all 22 biases →