FORGE Engineering

FORGE is an Autonomous Physical Systems Engineer that observes mechanical systems, builds evidence-backed physical models, tests engineering hypotheses, and verifies proposed improvements using deterministic physics. Today's multimodal AI can analyze an image and generate plausible engineering answers, but plausible is not the same as physically correct. FORGE addresses this gap by separating AI perception and reasoning from physical verification. In the current prototype, a user uploads an image of a mechanical transmission. Meta SAM 3.1 detects visual components, but FORGE deliberately treats these detections only as observations. An evidence boundary prevents visual confidence from becoming physical truth: properties that cannot be established from the image must be explicitly confirmed before engineering calculations are allowed. FORGE then constructs the physical topology and uses a deterministic Python physics engine to calculate and verify the system's behavior. Once a system is verified, the user can provide an engineering objective. Meta Muse Spark acts as a bounded engineering agent that proposes candidate redesigns. Each candidate is evaluated as a physical experiment by the deterministic engine. Failed results can become evidence for another attempt, while only a deterministic PASS can produce a VERIFIED design. FORGE does not ask AI whether an engineering solution works. It tests it. AI PROPOSES. PHYSICS VERIFIES.