Tensol has raised from Afore Capital, Carnegie Mellon and Graham Neubig (OpenHands).
Agents
The linking work, done between reviews.
Tensol’s agents tie each run back to the requirement it verifies, check that the rig was entitled to verify it, and assemble the package for the next review, so the review starts from a current picture instead of a rebuilt one.
Spacecraft shall retain ≥ 20% 30% battery charge through a 36-minute eclipse.
Derived requirements
- EPS-021Battery capacity ≥ 120 Wh at end of lifeNeeds revision
- ADCS-011Eclipse safe-mode draw ≤ 18 WNeeds revision
- TCS-008Heaters hold cells ≥ 5 °C in eclipseRe-check
- COM-004S-band downlink ≥ 2 MbpsUnaffected
Evidence
12 → 7 runs passing5 sign-offs stale2 test cases to re-authorWhat they do
Three jobs nobody has time for before a review
Each one used to be a spreadsheet pass the week before.
Linking
Every run, tied to what it verifies
A run lands and is matched to the test case it executed, the requirement that case verifies, and the channel that carries the result. The link is made from the data, not typed into a field.
tvac03_eclipse_044
61 channels · ingested in 38 sassert soc_min ≥ 30%Admissibility
A pass on the wrong rig doesn’t count
A run can pass and still verify nothing. Each result is checked against the level the requirement demands, and anything from a rig below it is held out of coverage instead of counted.
Eclipse depth of discharge
Review packages
The review package, assembled from the record
Before a gate, the agent gathers the evidence for every requirement in scope, marks what’s missing or stale, and hands the review a package that matches the record, not a deck someone rebuilt.
Review package
Five sign-offs went stale with rev 5 of SYS-014. They are listed for re-review, not dropped from the package.
See it on your data
Point an agent at one subsystem and watch the chain build.
Bring its requirements and a handful of representative runs.
Book a demo