what we refuse

We record judgment. We never let it decide.

Every mechanism here that judges writes a report a person reads. Nothing acts on its own score.

A person decides

Judgment is evidence, never authority. A tie is reported as a tie.

running

Your work never trains a model

No weight update, ever. What we learn is written down as text people can argue with.

running

Nothing is settled by counting

Not by vote, and not by a tally. We publish no figures without their population.

running

No number instead of evidence

No score, no dial, no percentage standing in for the thing. We show you the change.

running

We name what is unproven

Beside the claim, unasked. A page silent about it reads like a page that passed.

running

Nothing recorded is rewritten

Added to, never edited. A later reading argues with an earlier one; it does not replace it.

running

An exception is visible and narrow

Named, readable by you, and unable to widen quietly.

running

A check blocks only when it needs no judgment

If reading the finding takes an argument, it reports. It does not stop you.

running

No claim of measured superiority

We have not measured ourselves against anyone, and will not say we have.

running

A person publishes

We could automate the release. We decided a release is a person deciding.

refused

No loop that closes on itself

Nothing we ship feeds its own judgment back into the work with nobody in the path.

refused

No training on your work

Learning here means a corpus, not a model.

refused

A score can move for reasons that are not quality. So we report. We never act.

Nothing recorded is rewritten.

a reading is frozenWritten on the day it was written. A later one argues with it.
a retirement keeps its argumentWhat went, and why. The reasoning is the part worth keeping.
history moves forward onlyThe rule the company is named for. It was in the code first.