I built Stock Copilot as a solo PM who codes. It works. But I’ve hit a ceiling that more Cloud Functions won’t fix. The evolution is from a time-driven pipeline to an orchestrator-worker system where specialized agents pursue research angles in parallel. Here’s what that looks like — and the honest gap between today and that future.
A stroke patient opens the app to do cognitive exercises. Behind that simple moment, four things need to coordinate: which exercise to show, how hard to make it, whether the patient is struggling, and whether they should stop. Three services run concurrently, each owning different decisions.
Stock Copilot does five things every day — refresh prices, run analysis, generate audio briefings, push notifications. The naive approach couples all five into one blast radius. Here is how a time-driven pipeline with Firestore as the inter-stage contract solved it.
An iPadOS cognitive rehabilitation platform for stroke patients, built around offline-first sync and AI-orchestrated therapy sessions. Four role-shaped apps, one design system, HIPAA-aware from day one.
An AI-native trading workbench that turns plain-English questions into structured analysis across eight specialist services. iOS app, React PWA, and a model that never touches the trade button.