LapDelta

LapDelta replays real Formula 1 sessions in the browser: a to-scale track map, twenty cars at sixty frames a second, synchronised telemetry traces and a session-aware timing tower. It runs for roughly a pound a month.

It has no request-serving backend. That reads like a limitation I worked around, but it is the decision the rest of the design falls out of. A service that interpolates frames per client is a service you pay for per client, babysit, and scale. Removing it pushes the compute to the two places that are already free: once, offline, when a session is ingested, and then in the browser on the viewer's own machine.

What travels between them is a single precomputed file. An offline build in Python ingests a session and writes one artifact. S3 behind CloudFront serves it as immutable, brotli-compressed bytes. An Angular client with a Rust core compiled to WebAssembly downloads it once and renders everything from it. There is no /api, and no database in the request path.

The plate above comes from that same data, cut down to what a static page needs. The circuit is not a track map from anywhere. It is derived from position samples, the way the running system derives it, and the plate below shows a selection of circuits produced by that same code without any special-casing.

The full account of the geometry, the zero-copy data contract and what it costs to run is in the essay. The application itself is at lapdelta.app.