01 — Process
How I build, end to end.
Software engineering, AI automation, and product & UX run as one loop. AI agents don't replace the judgment — they remove the friction between the seams, so one engineer can ship like a team.
Click a node — see how AI fits in, and where I don't let it
Find the right problem
Discover
Talk to users, read the data, kill the ideas that don't survive contact. The job here is to be specific about who hurts and how much.
Where AI helps
- →Synthesize interview transcripts
- →Cluster support tickets by theme
- →Pull competitive teardowns
Where humans matter most
Picking which pain to chase is taste, not output.
02 — Tools I trust
The stack behind the loop.
I keep this list short on purpose. Tools earn their place by compressing time, not by being shiny.
01 · Plan
- Claude (Opus)
- Notion AI
- Granola
- Linear
02 · Design
- Figma
- Figma AI
- Origami
- Lottie
03 · Build
- Claude Code
- Cursor
- v0
- GitHub Copilot
04 · Ship
- Vercel
- GitHub Actions
- Sentry
- Statsig
05 · Operate
- Plausible
- PostHog
- Linear
- Slack
03 — Start a project
Have a problem worth solving?
I take on a small number of senior engineering & AI-automation consulting engagements. Best fits: AI-native MVPs, 0→1 product builds, hard-to-ship slices of larger products.