Now

Doing

• Defining AI-native hardware products. My part is model capability evaluation and the data pipeline: can a feature be built, how accurate is it in practice, what does each run cost.
• Wearing a recorder for weeks and using hundreds of hours of my own real audio to write exams for AI, to see what it actually remembers.
• This site just launched. Four topics to start: AI hardware, wearable health data, multimodal memory benchmarks, unit economics.

Paying attention to

• Which wearable health numbers can be trusted. Five devices on at once, and three different definitions of HRV.
• The unit economics of AI products: per-unit cost, gross margin, break-even, from day one.
• Small AI-powered businesses that get paid per outcome, where one-time effort earns for a long time.

Body

• Amateur rowing, on the water plus the erg every week, with strength and conditioning.
• Long-term trends over single days. One day's noise is bigger than the signal.

Contact

Start with the problem. Details in the footer.