The problem
My old cooldown routine was the same few drills every day, so I stopped doing it.
Why I have it
Huberman shows up on its own each evening with a cooldown built for the day I actually had, and it rotates through a library so it doesn't repeat itself.
What it does
- Reads the day's training and recovery, and what the gym programmed.
- Builds the session from a library of 25 drills, skipping the ones I've done recently.
- Says so when it can't tell whether I trained, and offers the other version.
Where it helped
- Live every evening since Aug 24, and on rest days since Aug 25.
- Never empty: if the model can't answer, the session still comes from the library.
- 6 regression tests pinned on it.
How it works
One message, step by step
Each step lights up in turn. The names are the real components in the code.
- Evening tickOnce a day, nobody has to ask.
- Read the dayTraining and recovery, from the vault.
- Drill library25 drills. Recent ones are skipped.
- The modelWrites the coaching, with a 90-second limit.
- Never-empty floorIf the model fails, the library session still goes out.
- TelegramSent. A missed send is retried the next minute.
Not built yet
- It's reached through rules and the evening tick; it isn't registered as a production agent yet.