THE METHOD

A result worth checking.

Tesven compares an expected outcome with independently read evidence. An agent’s success message is never enough on its own.

01. Define what should happen

A task needs observable conditions: a file with specific content, an updated value, or an expected state. Vague goals cannot become precise verdicts without defined checks.

02. Receive the result and evidence

Tesven receives a task attempt and access to authorized evidence. The agent’s claim is recorded as a claim, not accepted as proof.

03. Read the actual state independently

The configured verifier reads the authorized source and compares it with the expected condition. A trace or screenshot alone does not establish every possible outcome.

04. Record a scoped verdict

Each criterion has a result and evidence references. The outcome describes the conditions checked; it is not a guarantee about everything the agent did.

  • Verified: evidence supports the condition.
  • Failed: evidence shows the condition was not met.
  • Insufficient evidence: a reliable decision cannot be made.
  • Human review: ambiguity or conflicting evidence needs a person.
  • Not applicable: the condition does not apply to this attempt.

What if something is missing?

Unavailable access, incomplete evidence or an integrity failure must not become success. Tesven records an explicit non-verified state.

See the difference
Interactive demonstration · example data
Dovren

Update the website logo.

Task completed.

The agent’s claim.

Tesven4/4
  • Website reachable
  • Expected logo found
  • Deployment active
  • Change persisted after refresh
Verified

The evidence supports the expected change.

Illustration of the method. This is not a live run or an application screenshot.

AI does the work. Tesven verifies.

Get Tesven ↗