Confirmation bias is your co-founder now

When you build alone, every judgment call gets made by a committee of one. You pick the idea, you decide what counts as evidence, you grade your own test results. And the reviewer you report to has been compromised from day one — it wants the idea to be true. Psychologists call it confirmation bias: we search for what confirms our beliefs and rationalize away what doesn't. For a founder mid-idea, it's less an occasional glitch than a silent partner with veto power over the truth.

The easiest investor you'll ever pitch is you

The classic demonstration is Peter Wason's rule-discovery experiment from 1960: given the sequence 2-4-6 and asked to find the rule, people overwhelmingly propose triples that fit their hypothesis and almost never propose one designed to break it. Sixty-plus years of replication later, the pattern holds: we test our beliefs by looking for yes.

Founders are a spectacular case. Cooper, Woo and Dunkelberg surveyed 2,994 entrepreneurs and found 81% rated their own odds of success at 70% or better — a third put them at 100%. Asked about a business just like theirs run by someone else, only 39% were as generous. The gap between those two numbers is the co-founder talking. It isn't the market speaking; it's the wish.

How it quietly runs your validation

The bias doesn't announce itself. It shows up as process — reasonable-looking validation that's rigged without your noticing. You demo before you ask, so every question afterward is about your solution instead of their problem. You ask "would you use this?", a question social politeness answers for people. You leave interviews with what Rob Fitzpatrick calls bad data — compliments, hypotheticals, wishlists — and it all files neatly under "positive signal." You didn't run a test. You ran a showcase.

The receipts show up later, in the shutdown notes. When CB Insights analyzed hundreds of startup post-mortems, the top product-level reason was "no market need" — cited in 42% of failures. In many of those stories, a darker detail shows up: the disconfirming information existed before the product was built. Customers could have said the pain wasn't painful enough or the price wouldn't clear. The founders didn't ask — or asked in ways that could only return yes.

Cheap code made the bias cheaper too

There used to be a natural brake on all this: building was expensive, so at some point the cost of being wrong forced the question. Coding agents released that brake. When a working prototype costs a weekend, "let's just build it and see" feels like empiricism — but momentum is not evidence, and a growing repo confirms nothing except that code is easy now. We've written before about why ideas die late; confirmation bias is the reason they get to live that long in the first place. The cheaper building gets, the more the scarce skill shifts from writing code to judging whether it should be written.

Make it argue back

You can't uninstall the bias — it's part of how the machinery runs. What you can do is build the counterargument into your process, before your judgment gets a vote. Write the kill criteria first: before any test, put on paper what result would mean stop, so the goalposts can't drift once the data arrives. Ask about the past, not the future — "when did this problem last cost you something?" beats any hypothetical. Count commitments, not compliments: time given, money offered, an intro made. "Sounds cool" is worth nothing; a calendar invite is worth something. And design at least one test whose explicit job is to break the idea — if everything you run can only confirm, you're still playing 2-4-6.

Most of all: get a counterparty. What a real co-founder does — the useful kind — is not share your excitement. It's to hold the case against, out loud, while there's still time for the answer to be no.

Where Motriz fits

This is exactly the seat the co-founder in Motriz occupies. Its job isn't to share your excitement — it reasons through the product with you and keeps the scorecard: the evidence for, the evidence against, and the risks still open. At the checkpoint it recommends a call grounded in that scorecard — continue, strengthen the evidence, reshape, proceed to build, or stop on evidence — and the decision stays yours, recorded with its reasoning either way. The bias still lives in your head. It just no longer gets to grade its own homework.

Earlier in this series: the validation gap, how to validate an idea before you build it, and killing ideas early is a builder's superpower.