How I program deterministically around a non-deterministic LLM
Two production bugs from a legal AI platform, and why the fix for a non-deterministic LLM is engineering the boundary it reads, not writing a better prompt.
Read articleWriting
Notes from the work
What I have learned building AI-native systems, working with AI coding agents, and keeping production software reliable.
Stripe Connect escrow for a marketplace: the payout guards that actually hold
How I built escrow payouts on Stripe Connect for a marketplace, and which of the five guards against a double payout actually holds over time.
Read articleHow I test Postgres row-level security for tenant isolation
A row-level security test suite can pass and still prove nothing. How I built PI6's harness so a green run means a tenant really cannot read another tenant's rows.
Read articleHow I build local, open source agent software
The architecture I use for agent-powered tools that run on the user's own machine and credential, with the file system as the contract and no vendor lock, drawn from an open source tool called Marketing Ops.
Read articleHow I build MCP servers
How I built the multi-tenant MCP server for PI6: what I expose to an agent, the OAuth security around it, and how I dogfood it with an LLM.
Read articleWhat AI coding agents actually speed up
Notes from daily work with Claude Code on what genuinely gets faster, and where deciding what to build still takes experience.
Read article