// HEALTH · TRAINER IN POCKET
Beta-Ready · Pre-Launch
A personal trainer that lives on your phone. Daily check-ins, programmatic workouts, the kind of attention that used to cost $300 an hour — at the price of an app.
// The problem
The best workout-app market segment is built around two failure modes. The cheap end gives you a static plan and a counter. The expensive end gives you a human you'll never afford to keep. The middle — a coach who actually knows you, adjusts in real time, and fits in a pocket — quietly didn't exist.
What a great trainer does isn't programming sets and reps. It's noticing. They notice you slept badly. They notice the lift looked off. They notice you went two weeks without a check-in and back off the volume. The reps are a side-effect of the noticing.
"AI is finally good enough to do the noticing. The only question was whether we could put it in a check-in loop tight enough that someone would actually use it tomorrow morning."
// What it does
Check-ins
A short morning conversation: how did you sleep, what's sore, what's on the schedule. The coach reads your week, your last lift, your recovery — and adjusts before you walk into the gym.
Programming
Not a static 12-week template. The coach writes the next session based on what actually happened in the last one — load, RPE, recovery, mood. The plan stays in shape because the coach keeps re-shaping it.
Voice
The check-in voice is the load-bearing UX. We're rewriting it in something closer to a real coach — direct, warm, never AI-cheery. The phone says one thing the right way, not seven things the wrong way.
Native iOS
Loads instantly, works offline mid-set, syncs with Apple Health, renders beautifully on a watch. The phone is a tool, not a portal.
// Preview
// Where it is
Done
In flight
Next
// Get involved
Chala AI's private beta opens this summer to a small group. Join the waitlist for an early TestFlight invite.
// Under the hood
Tech stack
Native SwiftUI for iOS, on-device CoreML for the lightweight inference paths, server-side LLM for the heavier reasoning. Neon Postgres for state, Clerk for auth, Apple Health as the primary signal source.
This product writes to Drako DevOS's memory layer →
GitHub
Closed source. Reference implementations of the daily-companion + check-in patterns will land in the OSS Drako DevOS repo when it ships.
— COMING WITH OSS RELEASE
// Other projects