Most marketing copy is uncategorised feelings dressed up as fact. We do something different: every non-trivial sentence on this site gets a letter next to it. Here is what those letters mean and why we bothered inventing the habit.
The six classes, in one screen
There are six tags. We keep them short on purpose so they fit in the margin of a sentence without taking it over.
- A — empirical, in this session. A thing was measured or observed while the work was being done, and we can show you the trace. "The form posted and a row appeared in the leads table at 14:02." That is Class A if the row, the timestamp, and the request log all exist.
- B — code or inspection. The claim is supported by source you can read. "The retry logic backs off exponentially with a 30s ceiling" is Class B if the function is in the open and we link to it.
- C — configuration or integration. The claim depends on how the pieces are wired together right now. "Form submissions reach our database" is Class C, because it is true only as long as the route, the webhook, and the credentials remain configured the way they are configured today.
- E — expert citation. Somebody outside our shop, with a name and a paper, said this. We link to them. We do not pretend they endorsed us by being cited — citation is not endorsement.
- F — falsifier present. The claim comes with a stated way it could be proven wrong. If you can specify the test, you can attach this tag. If you cannot specify the test, you cannot attach this tag.
- U — unverified. We think it is probably true, we have not done the work to confirm. Honesty tax. We would rather say U than imply A.
If a sentence cannot earn any of these tags, we usually do not publish it.
The same sentence, three different classes
Take the line: "Our intake form delivers leads into a database that the agency reviews daily."
- Written as Class C, it is a configuration claim: the form, the webhook URL, the proxy, and the database are wired up today, and the daily review is on the calendar. That is honest. It is also fragile — if the webhook URL is rotated tomorrow and nobody updates the env var, the sentence stops being true. Class C names that fragility.
- Written as Class A, it would have to point at a specific submission that arrived and a specific review note that came back out. The claim shrinks to the event we can actually show.
- Written as Class U, it would mean we believe the pipeline works end to end but have not tested it since the last change. We would not publish a U on a load-bearing promise — we would go test it and upgrade the tag.
Same words, three different weights. The tag tells you which one you are reading.
Why we do this at all
Two reasons.
First, it forces us to do the work. You cannot tag a sentence Class B without linking to the code. You cannot tag it Class E without a real citation. The tag is a tiny audit that runs before the sentence ships.
Second, it gives you a way to push back. If we mark something Class F and you can show that our falsifier does not actually disprove the claim, we owe you a revision. If we mark something Class A and the artifact is missing, you can say so. The classes are not a shield — they are a handle you can grab.
There is a longer treatment of how this fits into the rest of the way we work on /trust-receipts-standard, and the underlying posture — receipts, transparency, partnership — lives on /transparency. If you want to see the practice in person, the place that happens is /workshop.
We will get sentences wrong. The tags are how we make that recoverable instead of embarrassing.
- Read the full standard: /trust-receipts-standard
- See the wider posture: /transparency
- Bring this into your own work: /workshop