One project in two parts for tuning car audio built around a sound processor. I gathered the approaches, documentation, instructions, libraries and experience (mine and friends’), and with AI’s help built the skill on that base: it makes the overall process deterministic while keeping AI flexible at every step. The skill follows a seven-phase algorithm, from intake to final listening, and works through many interacting factors at once: the car's acoustics, every speaker, crossovers, time delays, phase and equalisation. Real measurements and maths lead it to the best settings. The skill works in Claude, Codex and Antigravity.
TCC is a full desktop application built on top of the skill: multilingual, for macOS and Windows, with live measurements, the predicted result, the tuning plan and an AI chat panel. It works together with the skill and can use different AI models, directly or through the OMP tool.
In both, Claude calculates and proposes settings, a model from another vendor checks, critiques and corrects them, and the person listens and decides. Nothing changes the processor by itself.
7phases in one fixed tuning process
14major and minor releases of the skill since June 2026
4,300+automated tests in TCC
4languages; macOS and Windows
Results in 2026: my own car won five awards at EMMA and AYA championships, including two first places in qualifying rounds and third at the German EMMA Final. One user’s car took first place at the AYA Finale.