Writing
Notes on building with AI agents, measuring what they do, and the occasional thing from life.
- What I found reviewing an app my coding agent builtThe bugs that passed tests and looked fine in a diff: authorization checks that weren't where I assumed.Guard
- The 5-minute cron that billed me twiceTwo background jobs, two vendors, two surprise bills. What each one taught me about hobby infrastructure that never sleeps.Operate
- Designing air quality for a 1-bit displayHave Haze Not puts Singapore's air quality on a TRMNL e-ink panel. The interesting problems were all about which number to show.Build
- Giving Claude access to our household financesThe permission model behind an MCP server with real money behind it: off by default, writes behind a second switch, and rules the database enforces.Guard
- An agent waiting for permission that would never comeMy AI code reviewer ran for six hours doing nothing. In CI, every permission prompt is a hang.Operate
- 58% vs 90%: what measuring my own classifier changedAn ablation ladder over 450 of my family's real transactions, and the one product change it justified.Measure
- Navigating Our Financial Journey with SpendBotThe Ultimate AI-Powered Expense Tracking Companion
- Shedding Light on the Vaccination Regimen in SingaporeMy journey developing SG Vaccine Tracker