Day with AI: Fitness App on Supabase, Video Pipeline in 45 Minutes, and Cron Cleanup.
Fitness tracker migration from Convex to Supabase, complete video production pipeline as standalone repo, detailed training program PRD, and consolidation of 14 cron jobs. 28 commits, 15 sessions.

What I worked on
Wednesday. The day when "fix my data saving" turns into a full backend migration, a new video pipeline, and an infrastructure audit.
What I did
Fitness app — Supabase migrationMorning discovery: workout saving is broken in my personal fitness app. The existing backend setup was poorly configured. Solution? Complete migration to Supabase — new project, schema, integration. AI helped with the entire setup including MCP integration for direct database access. From "save doesn't work" to a functioning backend in one session.
Video production pipeline — standalone repoThis is where AI saves the most time. Complete video production pipeline — from scripting through components to onboarding — as a standalone repo with a universal profile that works for any creator. 21 visual components, human-in-the-loop review system, documentation. Tracked time from log: 8h+ of work → 45 minutes. That's not a typo.
Detailed training program PRDSpent the afternoon building a technical PRD for an advanced training program. AI went through complete exercise research — analyzing each exercise, volume analysis by muscle groups, identifying 8 specific errors in the original plan. The research produced a structured document with evidence-based recommendations. Work that would take a trainer a week, done in an afternoon.
Cron consolidation14 cron jobs in LaunchAgents — some running, some not, some redundant. Did a complete audit, explored new Claude Code Routines as a potential replacement, and added structure. Infrastructure maintenance nobody wants to do, but without it the system falls apart.
Rook cleanup + cold email pipelineMorning routine — syncing with the AI social media assistant, cold email pipeline generating leads, YouTube digest pulling new content. Automation running in the background.
Time: AI vs without AI
| Task | Without AI | With AI |
|---|
| Fitness app — backend migration | 2-3 days | ~2h |
|---|---|---|
| Video pipeline — standalone repo | 8h+ | ~45min |
| Training PRD + exercise research | 1 week | ~3h |
| Cron audit + consolidation | 1 day | ~1h |
| Cold email + social media sync | 3-4h | ~20min |
| Total | ~2 weeks | ~7h |
What I learned
- Backend migration with AI is surprisingly painless — MCP integration means AI has direct database access and can validate schemas in real time
- Video pipeline shows where AI excels most: repetitive creative work with clear structure. 21 components in 45 minutes would take a week manually
- Exercise research is a perfect AI use case — it processes hundreds of studies and returns a structured verdict. A trainer would do it from memory, AI does it from data
- 14 cron jobs is too many. Consolidation and regular audits aren't sexy, but without them infrastructure becomes unmaintainable
Interested in the article?
Let's discuss what this kind of automation can do in your company.
Free consultation