Latest
FixOurHood
Spot an issue, snap a photo, AI writes the report, city gets notified.
fixourhood.ca ↗Building for results.
I pair Ruby on Rails with AI agents to ship production apps at startup speed. Full product builds, from idea to live users.
Right now
Building an AI-powered Rails agent, think v0 and Lovable, but for Ruby on Rails. One prompt to a full, production-ready app with RapidRails UI components, Hotwire, and Tailwind. Shipping soon.
Also maintaining RapidRails (UI component library for Rails) and FixOurHood (civic reporting for Toronto neighborhoods).
What I do
I build products.
I started in design, moved to frontend, and landed on the thing I care about most: building products that people actually use. Not prototypes. Not demos. Real products with real customers.
Everything I build runs on my own Ruby on Rails stack and I pair it with AI tools like Claude Code to ship at a pace that used to require entire teams. FixOurHood was built 90% with AI-assisted development and is live in production. That's the workflow now: deep problem domain knowledge and developer expertise paired with AI speed.
My craft
I write code — with AI.
I got my start as a designer, but became obsessed with the gap between beautiful mockups and real, working software. That led to years of frontend & backend development, then UI component architecture, and now building full products from scratch.
Today I work with AI as a development partner. Claude Code, cursor-based workflows, agentic tools, I use them daily to write, review, and ship code faster. Not to replace thinking, but to multiply output. The result: I ship products in weeks that used to take months.
Work
Products & Projects
RapidRails
Production ready UI component library for built for Rails developers and the Rails ecosystem.
rapidrails.cc ↗ShipIt.day
Accountability tool for builders. Track your progress publicly and stay honest about what you are shipping.
shipit.day ↗One developer with Rails and AI can build what used to take a team. Simple solutions beat clever ones. The best tools feel obvious once you use them.
Writing
Latest Posts
Context window economics for AI agents
Every tool, MCP server, and progress file you add to an agent costs tokens. Most teams don't measure this. Here's how to think about context allocation.
Where verification actually belongs in agent harnesses
Anthropic says use MCP Puppeteer for screenshots. Geoffrey Huntley says keep verification outside the loop. Who's right? Both, but for different reasons.
7 Patterns for long-running agent harnesses
What I learned applying Anthropic's 7 harness patterns to my agent build system. The easy ones don't matter much — the hard ones are what separate shipping from spinning.
Get in touch
Let's build something
Need a Rails developer with UI expertise who ships fast with AI-augmented workflows? I help businesses go from concept to production, in weeks, not months.
Currently booking for Q2 2026 and beyond