Plan your next bin.

The coach reads your loaded tune and datalogs, bins every WOT sample onto the WGDC grid, and asks an N55-aware model for the highest-impact, knock-safe changes for your next bin — ranked, with the data behind each one. It only ever suggests changes the safety rules allow. Runs free on Cloudflare with no API key out of the box — bring your own Anthropic key for Opus or Sonnet.

1 · Datalogs

2 · Tune

Load your XDF and bin on the Tune tab and they carry over here — the coach reads your WGDC base, timing and boost tables straight from the bin. Optional.

WGDC base override

Paste a 20×16 base to override the one from the bin (or if no bin is loaded).

3 · Model

runs free on Cloudflare Workers AI — no API key needed

No key needed

This model runs free on Cloudflare Workers AI via the deployment’s AI binding — nothing goes to Anthropic and no API key is required. Switch to Opus or Sonnet for the deepest reasoning (needs a key).

0 logs0 WOT samples0 cells with database —0 tune tables0 safe presets

Preview what gets sent

# NEXT-BIN COACHING REQUEST

## Vehicle & tune
BMW N55 (single twin-scroll turbo, PWG), MEVD17.2 ECU, hybrid turbo, 100 RON fuel. WGDC base is a 20×16 feedforward grid — rows = boost target (PSI), columns = RPM.

## Datalogs
No datalogs loaded — reason from the current tune tables only, and say what logs you would need to go further.

## Your task
Using ONLY the data above, recommend the highest-impact, SAFEST changes for the NEXT bin. Rank them. For each: the table + exact cell(s) (PSI × RPM), current → suggested value, the data that justifies it (mean P-Factor, boost error, N, knock), a confidence level tied to sample count, and a one-line safety note. Obey every hard safety rule. If the data is insufficient for a given change, say so instead of guessing. Finish with a paste-ready copy block (updated WGDC base TSV or a per-cell delta list).

Load at least one datalog (or a tune with tables) to begin.