Harness Engineering — Visual 1 of 3

A Principle Is Not a Harness

Step 1 — The underlying principle: Fleming's Left Hand Rule

F = Force / Motion (thumb points up) B = Magnetic field (index points right) I = Current (middle → toward you) Lorentz force law F = IL × B context-free · always true

Three perpendicular vectors: field (B), current (I), force (F). This rule applies in a lab bench, a rail gun, or deep space. It is context-free physics — it tells you what will happen when current flows through a magnetic field.

Step 2 — The harness: a DC motor wraps the principle

N pole S pole B → commutator output shaft Harness components ① stator + magnets provide the B field ② wound armature current-carrying coil in field ③ commutator ★ reverses I each half-turn ④ brushes sliding electrical contact ⑤ housing + shaft transmit + contain output

The ★ commutator is the critical insight. Without it, the coil would oscillate back and forth (a twitch). The commutator reverses current every half-rotation so torque always pushes the same way — producing continuous, controlled rotation. That's what a harness does: it makes the principle keep going productively.

Step 3 — The same structure in machine learning

DC Motor LLM Agent F = IL × B Lorentz force law harness • stator + magnets • wound armature • commutator ★ • brushes + housing continuous rotation drill · fan · EV drivetrain transformer + attention the architecture as principle harness • system prompt + context • tool definitions • execution loop ★ • memory + guardrails autonomous agent action code · research · decisions same structure

The ★ execution loop in an AI agent does what the commutator does in a motor: it keeps things turning in the right direction. At each step — call model → execute tools → feed results back → repeat — it prevents the system from oscillating and ensures the principle produces useful, directed output.

1 / 3