Workbench
2026- Product Design
- Design Engineering
Lorem ipsum dolor sit amet consectetur. In feugiat nam elementum volutpat. Sit id quis amet tortor. Orci tempus sit diam neque donec.
- Daily driverfor my own workflow
- Zero depsruns offline

Lorem ipsum dolor sit amet consectetur. In feugiat nam elementum volutpat. Sit id quis amet tortor. Orci tempus sit diam neque donec.

Lorem ipsum dolor sit amet consectetur. In feugiat nam elementum volutpat. Sit id quis amet tortor. Orci tempus sit diam neque donec.
Lorem ipsum dolor sit amet consectetur. In feugiat nam elementum volutpat. Sit id quis amet tortor. Orci tempus sit diam neque donec.

A distributed warehouse network enabling same-day delivery SLAs.

I work on problems that show up messy.
Most of what I’ve built is software people use all day to get their job done — moving inventory, raising money, running an operation. The hard part is rarely how it looks. It’s that people have already built habits around what’s broken, and the team has strong opinions about a process nobody can fully explain.
I started as a developer and never really stopped. Building a working prototype turns a question into an answer the same day, and engineers get something real to react to.
I’ve also shipped my own things end to end — web and mobile apps, VS Code extensions, AI agents, even a 3D rendering engine, and much more. It keeps me honest about what’s actually hard to build versus what just looks hard.
When I’m not building, I’m usually climbing, hiking, or exploring somewhere new.
It’s part of the day-to-day now — not a separate tool John switches into. A few places it earns its keep for him:
Exploring alternatives
He’ll ask for six variations of a layout or a piece of copy, pick two, throw out four. Cheap to explore, easy to discard.
Pushing past first-draft framings
His own or the team’s. When something feels safe, he uses AI to force different angles before locking in.
Compressing the boring parts
Placeholder data, edge-case copy, quick throwaway prototypes. What used to take an afternoon lands in under an hour.
He still makes the calls — AI just widens what he can plausibly explore in a given afternoon.