I wear a WHOOP. I've coached people on movement and sleep for many years and I still can't answer that question for myself. The algorithm is locked. You get a number, you trust it, you stop there.
When we built Open Wearables, we decided the scoring layer should work differently. Sleep Score and Resilience Score shipped in v0.5 - every coefficient, every threshold, every weighting is in the repo and you can fork them, tune for endurance athletes or elder care or clinical populations. Moreover, you run them on your own infrastructure and the same algorithms feed the MCP layer so AI coaching can cite the actual data behind a recommendation instead of approximating.
From the very first version, we ran it on itself. The MCP kept our code quality high from day one, which meant the agent always had clean, well-structured code to build on. The feedback loop was tight enough that we could ship feature-complete in a fraction of the time we'd expected.
Then we hit a wall. Performance wasn't where it needed to be, and it was baked into our choice of language (Python). There was no patching our way out of it.
I meant it. I had done it before. I know what hardware costs, not just in money, but in decisions you make at 2am about components that may or may not arrive, about inventory that ties up capital for months before a single unit ships. When I moved into SaaS, the relief was real. Software scales. Software does not sit in a warehouse.
