The physical recorder now fails closed.
The bounded SocketCAN capture path verifies listen-only mode before recording and preserves a compact safety-evidence sidecar with every session.
Season One · Building in public
Find the signal. Prove the change.
A passive-first vehicle observability project for curious owners, makers, and researchers who would rather measure than guess.
The learning loop
RadarWrench is exploring a more approachable path from raw vehicle traffic to evidence you can understand and verify.
Listen to what the vehicle already says and preserve the raw evidence.
Perform one deliberate action and place a precise marker in the timeline.
Rank what changed, repeat the experiment, and build confidence from proof.
Measured, not imagined
The first experiment asks a simple thing: when a known action happens, which bits change consistently enough to deserve a closer look?
The first rule
Season One begins with passive observation and isolated bench experiments. It does not begin with arbitrary commands on a live vehicle.
Field notes
The project is early. The receipts are real.
The bounded SocketCAN capture path verifies listen-only mode before recording and preserves a compact safety-evidence sidecar with every session.
Two isolated Raspberry Pi channels exchange known frames at production bitrates with clean counters, stable interface names, and no power warnings.
A galvanically isolated PCAN-USB FD captures known classic and FD traffic in kernel-reported listen-only mode while transmit counters remain unchanged.
The first end-to-end experiment preserves raw events, replays them, and ranks the intended bit consistently across three trials.
Follow the proof
The bench is proven and the receive-only guardrails are in place. The next result will come from a bounded, parked-vehicle observation with the engine off.
Say hello