▮ RESEARCH & EVIDENCE

Measured, with stated limits.

Gammatic proves separate claims separately. Each result below states what it demonstrates and what it does not. It is the same discipline the product applies to enterprise claims.

FLAGSHIP CASE STUDY: FACTORY PRE-DESIGN

10 engineers · 4 months

1 day.

The thesis case study delivered a complete factory pre-design, from requirements to a documented, internally consistent concept. Work of that scope is conventionally planned for a ten-engineer team over roughly four months. Under Gammatic governance, the multi-agent system completed it in a single working day.

SINGLE CASE STUDY FROM THE THESIS. THE CONVENTIONAL BASELINE IS A STANDARD ENGINEERING ESTIMATE. NOT A GENERAL PRODUCTIVITY GUARANTEE.

100%

Of controlled thesis runs ended on an external, auditable stopping event. Without the control layer, most runs hit their turn limit with no decision at all.

DEMONSTRATES A MECHANISM-LEVEL DIFFERENCE IN STOPPING BEHAVIOR.

Every claim

Leaves the system with a status: verified against its source, explicitly marked unverified, recomputed under deterministic rules, or rejected. Unsupported numbers do not pass as fact.

PROTOTYPE BEHAVIOR OBSERVED IN STORED RUNS.

SOURCE: “DESIGNING UNDER UNCERTAINTY: THE GAMMATIC MULTI-AGENT FRAMEWORK,” B.SC. THESIS, POLITECNICO DI TORINO, 2026 · INTERNAL SYSTEM REPORT, JULY 2026

▮ CLAIM DISCIPLINE

We type our own claims.

A system built to grade claims should expect its own statements to be graded. Everything Aurvionics says publicly carries an epistemic status and a basis.

Status Claim Basis

Supported A complete factory pre-design case study was produced in one working day under Gammatic governance. THESIS CASE STUDY

Supported Working prototype in a simulated enterprise environment. CODE + STORED RUNS

Supported The prototype reconciles selected numerical claims against retained source values and domain rules. CODE + STORED RUNS

Observed Stored runs include examples of reasoned rejection of a false user premise. RECORDED EXAMPLES

Observed The prototype declines calculations when the required data grain does not exist. RECORDED EXAMPLES

Vision Longitudinal institutional learning is part of the frozen target vision. DESIGN INTENT

EVERY PUBLIC STATEMENT CARRIES ITS STATUS. STRONGER CLAIMS FOLLOW STRONGER EVIDENCE.

▮ GOVERNANCE ALIGNMENT

Designed to map to recognized frameworks.

Gammatic's evidence and governance architecture is designed so future implementations can be mapped to the NIST AI Risk Management Framework, ISO/IEC 42001, and W3C PROV provenance concepts. Formal alignment work is scoped as part of enterprise deployment, not assumed in advance.

▮ EVALUATE THE SYSTEM

Request a briefing See the technology

THE FULL WHITEPAPER AND EVALUATION PROGRAM ARE AVAILABLE UNDER NDA.