Last Week Last week I recorded the update from a parking lot while the houseguests I was hosting rotated through the city. I used the downtime to map Google Play’s closed testing requirements, built a checklist for the fourteen testers I need to keep the track alive, and rewrote my dogfooding playbook so the first bug hunts have scripts instead of improvisation. Most of my energy went into people logistics: scheduling those friends and family members for thirty-minute screen-sharing sessions, documenting how to install from the Play Console instead of sideloading, and figuring out how to capture their feedback without asking them to learn SQL....

Last Week Last week I finally coaxed the Supabase cron job into draining the email outbox without manual babysitting. I replayed 500 queued notifications, watched the Kotlin client hand them off to Postgres, and confirmed Resend responds within the same edge invocation. With that loop stabilized the admin dashboard now reflects message success within seconds, and I officially declared the infrastructure feature complete. I did try to finish the Resend webhook, but Supabase keeps rejecting the callback with a 401 even after I disabled JWT enforcement....

Last Week Last week I finally pried open the Supabase email pipeline that had been stuck in limbo. I rewired the outbox drainer so it writes to a dedicated log table before and after each batch, captured the payload that Resend expects, and added defensive retries around the queue reader. Most of that effort went into proving the cron job could wake the Edge Function reliably; I ended up setting up a cron_debug_log table, replaying scheduled jobs, and verifying Supabase’s pg_cron extension was actually firing once a minute....

The Great Migration After months of using Claude Code as my primary AI coding assistant, I’m switching to Codex CLI. This isn’t clickbait - it’s a fundamental shift in my development workflow driven by tangible differences in model behavior, cost efficiency, and most importantly, trust in the tool’s ability to follow instructions. The AI landscape changes rapidly. Until recently, Claude Code and Claude models dominated agentic programming tasks. If you could afford it, Claude was the obvious choice....

Last Week Successfully set up the email outbox infrastructure in the database - a queue table that holds all emails waiting to be sent. Connected the app to this outbox so guest notification emails now properly queue instead of attempting immediate delivery. The architecture is solid: emails queue, the drainer function works when manually invoked, but the cron job automation isn’t triggering. The Cron Job Mystery The email infrastructure is 90% complete....

Last Week Recording outside as summer comes to an end! Last week was productive - I successfully set up Kermit the Log and used it to diagnose a phantom UI error. The database insert was succeeding, but the frontend was showing failure. Turns out the post-insert fetch operation had a parsing error that was masquerading as an insert failure. Fixed the error handling to report these separately and resolved the parsing issue itself....

Taking a Break from Shokken Not every day is spent writing Kotlin and wrestling with subagents. Sometimes you need to explore other creative outlets. Today, I’m diving into something completely different: training LoRAs using Ostris’s AI toolkit for image and video generation models. When I first opened this tool, the wall of configuration options was overwhelming. Every toggle, every number field seemed important, but I had no idea what any of them actually did....

Last Week The presentation layer hookup to the backend is complete but riddled with a perplexing error: database operations succeed but the UI reports failure. This phantom error led to implementing proper logging infrastructure – enter Kermit the Log from TouchLabs. The implementation journey also revealed important lessons about when subagents help versus hinder development progress. The Phantom Error and the Need for Logging The error manifests in a maddening way: add something through the UI, the database operation succeeds, the record appears in the database, the operation returns successful – yet the UI stubbornly displays “Operation failed....

Real Development, Unfiltered This is 1.8 hours of unedited, raw development footage showing exactly how I work with Claude Code on the Shokken app. No cuts, no edits, just the real workflow with all its iterations, mistakes, and eventual victories. Watch as we debug a deceptively simple phone validation issue that reveals deeper insights about AI-assisted development. The Setup The development environment showcases several key components: JetBrains Gateway for remote development (my machine is just a thin client) Ubuntu home server hosting the entire development environment Physical Pixel tablet for testing (emulators don’t work with remote development) Claude Code with Opus model ($200/month subscription) Seven specialized subagents for different development tasks The project is Shokken, a Kotlin Multiplatform app using Compose Multiplatform, targeting Android, iOS, desktop, and web platforms with strict clean architecture and 95% test coverage requirements....

Last Week The transition from mock data to real backend integration began in earnest. The home dashboard, which had been displaying carefully crafted fake data, started connecting to the actual Supabase repository through properly architected domain and data layers. Connecting the Layers The waitlist management features are now fully operational – users can view their created waitlists and add new ones through a complete vertical slice of the architecture. The guest list details pane is partially implemented, with some testing challenges arising from the transition from mock to real implementation....