Team 59

Evergreen

Evergreen is a personal training agent for busy people who want to work out more: it reads Apple Health and your calendar, plans workouts around the time you actually have, and fixes the plan on its own when life gets in the way

Check your own project

Community votes recognize the build and do not affect official judging.

Review the project

Start with the source code, then open the demo or video if available.

View GitHub repository
Project description
Evergreen is an iOS training agent that plans around your real life and acts end to end, with an audit trail behind every decision. How it works: it reads Apple Health (runs, heart rate, training load) and your calendar, then schedules the week's workouts into the time you actually have, each with a visible reason. When something breaks the plan it acts without a tap: a meeting lands on a run, so it moves the run to the next open slot, with one-tap Undo. Rain is forecast, so it finds a $15 gym day pass nearby and books it once you approve. Spending always needs your OK. Shopping: when a runner's shoes hit 392 miles nine days before a race, our recommendation harness on Vercel uses Nimble to research real shoes on expert review sites. Hard rules run in code (wide toe box, no carbon plate, arrives in time to break in). Superlinked's SIE ranks the shortlist with a Qwen reranker, and a Qwen model picks one and explains why every other shoe lost. Senso supplies cited passages live, so every claim has a source. Every recommendation is written to a SingleStore audit log, and guardrails fall back to a safe deterministic pick if anything fails. Memory and privacy: a memory agent runs on the phone in JavaScriptCore, rolling your history into a model of how you actually train. Workouts, memory updates and approved purchases go to Amplitude with opaque hashed ids. No names or health readings leave the phone.
Tools used