A local model may improve the wording. It may not change the claim.

Truemend writes plain English about a finding’s history. Those sentences are written by code from your own run history, never by a model.

Rewording is optional, runs on your hardware, and is off by default.

Decided before any model sees it.

Written by the engine
This finding first appeared 4 months ago and is still open.

Your run history stays on your machine. The engine reads it and writes the sentence. If rewording is enabled, that finished sentence is the only thing the model receives, and one sentence is the only thing it may return.

Three things it cannot do.

Cannot invent

It is never given a number, a count, or a finding. There is nothing in front of it to get wrong.

Cannot overstate

A version that would say something stronger, weaker, or other than the engine’s sentence is discarded, and the original ships.

Cannot phone home

The model runs on your own hardware. An address that is not local warns you, because it would mean text leaving the host.

20 / 20
Adversarial rewrites rejected

Twenty rewrites that alter the claim in the ways a model plausibly would were replayed against the guarantee: reversing subject and object, hardening a bounded count, adding a cause that was never stated. All twenty were rejected. Eight faithful rewrites were accepted.

One of the twenty is real. During a benchmark a model turned a sentence about a finding into a sentence about the runs: same words, same number, nothing dropped, and the claim reversed.

Disk, and nothing else.

No account, no API key, no service to reach.

Disk
~5 GB

The model weights, downloaded once. About 20 MB for the local server.

Default
Off

Output is byte-identical for anyone who never enables it.

Hardware
CPU is fine

A GPU is roughly six times faster at identical output. Speed only, not better wording.

Network
None

Nothing leaves the machine. There is no service to sign up for.

Worth saying plainly.

×
Not AI detection.

Findings come from analysis of your code. No model is involved in deciding what is wrong.

×
Not AI fixing.

Fixes are mechanical transforms. This feature touches one thing: the wording of history sentences.

×
Not a summariser.

It does not read your findings and tell you about them. It receives one finished sentence and returns one sentence.