ᛊ SOL · Independent Agentic Assurance / Adversarial Hardening

An independent verdict for the outputs of agentic systems.

From outside the producer boundary, SOL falsifies and audits specific artefacts, outputs and behaviours produced by agentic and post-agentic systems. Its first-class product is an independent attestation, not a test-run report.

Describe your problem

01SOL

Nemo iudex in causa sua.

No one may be judge in their own cause. A producer may and should test its own work, but an independent verdict must originate outside the producer boundary.

Flagship project · Gauntlet

Gauntlet tests a system from the outside.

It examines an exactly identified artefact, output or behaviour of an agentic system without taking over the producer role. The verdict is bound to a version, scope, suite and limitations—not a general marketing promise.

How it works

the request binds the evaluation to an exact target and scope
Gauntlet selects deterministic and adversarial probes for the claim
producer evidence remains separate from the independent verdict
the result includes findings, a minimal reproducer and retest criteria

Who it is for

For teams that need to trust a result, not a declaration.

For builders, operators and buyers of agentic and post-agentic systems that need evidence from outside the producer team before a release, deployment or procurement decision.

Availability

The tooling works. A repeatable pilot service is being prepared.

Versioned probes, contracts and a public audit example already exist, but the repeatable execution plane and independent signing, expiry and revocation remain under development. The waitlist is for preparing the first pilot evaluations, not issuing ready-made certificates.

In the form, write “Gauntlet” and describe the system, artefact or claim that should be tested.

Join the Gauntlet waitlist