Absent
The registry answered. Nothing was there - a finding you can act on.
THE METHOD
Each stage is real, each one can stop and say why, and every claim it produces carries the source it came from.
01
Real registries answer before anything is built - npm, PyPI and GitHub. The verdict is sealed and hashed before any outcome is known, so it can never be edited to look smarter later.
✦ Proceed or fail, sealed02
A Proceed becomes a schema-validated plan: screens, features, a data model, a source of truth - and the kill criteria you commit to before you build.
✦ A plan you can priceSCOPE
03
The build is rented infrastructure - the commodity, not our edge. Your source is never stored, only artefact hashes, and “done” means the URL verifiably responds.
✦ A live, verifiable URL04
Real telemetry against the bets you set in advance. Empty reads say “no data yet” rather than showing an invented number, and every finding appends to the record.
✦ Evidence over timeRETENTION
THE CARDINAL RULE
Every other tool paints one grey dash for both. This single distinction is the difference between evidence and a guess.
The registry answered. Nothing was there - a finding you can act on.
We could not ask. We learned nothing. Our failure, never your evidence.
Not promised in copy. Built into the system so it survives every future change.
The type system forbids it, and a test fails to build if it tries.
Only artefact hashes - invariant is enforced, not promised.
Security scans must clear, and the URL must verifiably respond.
Invariant: a self-reported figure is an observation, never blended in.
A run does not always end in a product. That is the point.
The evidence held. It shipped, and the record shows how.
The evidence said stop. That is a result, not a failure to hide.
Something broke. We say what, and where - never silently.
Validating an idea is free. The registries answer in about forty seconds.