Proof that translated systems still behave.
Strust replays the same work through the source and translated systems. If behavior moves, it shows the first decision that changed and the smallest patch that puts it back.
The source run becomes the specification.
Strust keeps only the parts of the call stack that shaped the outcome. Those thin execution slices are rebuilt into a replayable reference, so the translated system is judged by what the source actually did.
No guessed test cases. No translated code grading itself.
01
Capture the run
Record the calls, values, branches, state changes, and outputs that produced the source result.
02
Rebuild the behavior
Layer those execution slices into a reference the translated system can replay.
03
Isolate the change
Stop at the first slice that no longer fits, then keep the patch attached to that evidence.
Every confirmed mismatch becomes a permanent check.
Once a finding is approved, the next attempt inherits it. Every later migration wave starts with everything the system has already learned.
Observe once · keep forever
Conversion evidence
Bring us a conversion. We'll bring back proof.
A clear record of what stayed equivalent, what changed, and what a human reviewer still needs to decide.