Emergentism
[I] Portable practice · The method, on its own

How to check what you believe

This page has no philosophy in it. It is a habit for keeping track of how well you actually know the things you act on — and it works whether or not you agree with a single other page on this site. If you take one thing from here, take this.

  1. Write the claim as one sentence.

    If you can't get it into a sentence, you don't have a claim yet — you have a mood. That's fine, but label it as one.

  2. Label how well you actually know it.

    Not how strongly you feel it. How you came to know it.

  3. Write down what would prove it wrong.

    Be specific enough that someone else could go and check. If nothing could prove it wrong, it isn't a claim about the world — and you should stop spending decisions on it.

  4. When it's proven wrong, write the date beside it and keep it.

    Don't delete it. The list of things you were wrong about, with dates, is the only real evidence that your other beliefs have been tested at all.

The four labels

ProvenFollows by argument or arithmetic. Someone can check it on paper.
MeasuredWe looked. Here is the number, the date, and how it was collected.
AssumedLoad-bearing, not checked. We're building on it anyway — knowingly.
GuessedA hunch. It may be a good one. It is still a hunch.
The one rule that makes it work: a label may never quietly move up. A guess doesn't become proven because it's been repeated for a year, because it's load-bearing now, or because someone senior said it. Moving a label up requires new evidence and a date.

A worked example

Nothing to do with this site. A team is losing customers.

Claim "Customers leave because our price is too high."
Label Guessed — it came from three loud complaints and one bad week.
Would disprove it Interview 20 customers who left. If fewer than half raise price unprompted, the claim is wrong.
What happened 4 of 20 mentioned price. 13 mentioned a missing integration.
Died 14 March — kept on the list.
What it cost A discount plan was already on the roadmap. It would have lowered revenue and fixed nothing.

Notice what the habit did. It didn't make the team smarter — it made one cheap test obvious, and it caught a guess that had been quietly promoted to reason we're building things. That promotion is the failure mode. It happens silently, and labels are how you catch it.

Notice also what stayed. The dead claim is still written down, with its date. Next year, when someone says "we've always known it was the integrations," the record says otherwise — and says exactly when it was learned, and what believing the other thing cost.

That's the whole method

Four steps, four labels, one rule. It costs about ten minutes per claim and it works on product decisions, hiring beliefs, health advice, and arguments with yourself.

You can stop here. Nothing else on this site is required for it to work, and if you use only this page, it has done its job.

We hold ourselves to this. Our own list of dead claims is public, dated, and includes ones that cost us a great deal. If you'd rather see the ideas than the method, start plainly — or leave, which is a real option and always has been.