For the past 6 weeks I've been logging everything — workouts, macros, weight, sleep, mood — into a single profile that an AI coach reads each session. Trying to figure out if structured N=1 tracking with AI feedback actually changes behavior or just becomes another habit that decays.
Setup:
- Body stats → calculated maintenance (Mifflin-St Jeor) → daily macro target
- Workouts auto-tagged by muscle group, fed back as 7-day rolling exposure
- AI keeps the last 20 conversations as context. System prompt has my last 7-day calorie average and current streak
- Free Haiku 4.5 for quick chat, Sonnet 4.6 only when reasoning load is heavy
Findings so far (n=1, take with salt):
- The single biggest behavior change came from the AI calling me out by name on missed targets. Generic ChatGPT never did this — it didn't have the data
- Streak gamification stops working around day 21 once it becomes habitual
- I needed about 14 days of data before the recommendations stopped being generic
- The micro-frustration of logging actually helped — friction made me think before eating, not after
Open questions for the sub:
- Anyone tracking similar with a different AI / memory setup?
- What's your baseline window for trusting an N=1 trend (4 weeks, 8, 12)?
- Did anyone find a metric that mattered more than they expected?
(Built the wrapper myself — happy to share if anyone wants to try the same tracking setup, drop a comment.)