⸻ Vancouver, BC
I build developer tools. Right now that's Hookie and RSC Boundary, after 14+ years shipping product.
My first real job as an engineer was building marketing websites and newsletters at Backflip in Paris. Within a year I was the only engineer on a B2B ecommerce system a manufacturing company ran on: UI, backend, and the ops when checkout or inventory broke. The problems rarely arrived fully specified, and there was no one to hand them to. I found out I liked that. Owning the product end to end while the problem is still fuzzy became the thing I looked for in everything after.
It's why I built Liftmap at WiderFunnel, a CRO SaaS, when implementation alone stopped being enough. It's why I said yes to being CTO of Speakbox (Techstars W21) while working on TELUS's API team, two customer-facing apps and every architecture decision behind them. And it's why I ended up in developer experience at TELUS: the fuzziest problems weren't in the product anymore, they were in how engineers shipped it. Four years there. A self-service deploy experience on Cloud Run took time-to-production from weeks to under a day. GitHub Actions workflows spread to 50+ teams. An AI-assisted incident workflow produced 500+ reports in its first two weeks. Different products, same job: find the fuzzy problem, own it until it isn't.

Vancouver, BC
developer toolOpen-source webhook development tool. Receive, inspect, and relay webhook events locally. Go, Next.js, Redis Streams, gRPC. 3 releases.
open-sourceOpen-source devtool for Next.js and TanStack Start. Visualize React Server vs Client Component boundaries in the App Router. Zero production cost; agent skill for Cursor and Claude Code.
Techstars '21Mental health SaaS platform. Co-founded as CTO; Techstars W21.
How I secured the Hookie CLI release pipeline with OIDC trusted publishing, staged npm releases behind 2FA, and immutable GitHub release assets.
Notes from building Hookie: API-shaped commands are useful, but AI-native dev workflows need CLI commands that own the whole task.
Agent-loop discourse sells parallel code output. For solo builders the binding constraint is usually demand, and a useful receipt needs cost, artifact, and a user-visible before/after.
The industry says AI agents will write most of the code. Senior engineering interviews still gate on whether you can produce a graph traversal cold in 45 minutes. A checkpoint piece, four months into the job search, on the gap between the two.
Two albums on the streaming platforms. I write and record when the IDE is closed.
Training the Afro-Brazilian art of movement, music, and a fair bit of falling over.