Writing
Field notes and technical essays on developer platforms, agent context, system governance, and applied AI. The details cover design-system distribution, model boundaries, frontend architecture, and the tradeoffs that do not fit in a README.
Featured
June 2026 · 10 min read
Success Was the Incident
Adoption turned a five-minute browser cache into a budget problem. Splitting it into three caches with three owners, hashing dependencies behind stable component files, and a finops partnership cut request volume to a third of its peak.
- Platform Engineering
- System Design
- Design Systems
June 2026 · 13 min read
From Snippets to Shadow DOM
Why a jQuery-era design system splintered into forks and customized local versions, how Shadow DOM moved governance into the component boundary, and what it took to earn adoption from a thousand engineers.
- Design Systems
- Platform Engineering
- Technical Leadership
May 2026 · 9 min read
Teaching AI Agents to Use a Design System
A practical look at constraining AI-generated frontend code with design-system primitives, agent instructions, and repeatable validation.
- Design Systems
- AI Tooling
- Frontend Platform
All writing
Context Is a Budget
July 2026 · 7 min read
Context engineering as a budgeting problem: allowlists at the prompt boundary, cost claims verified against usage data, and instruction files kept thin enough to stay true.
Agentic Software Needs a Design System
July 2026 · 9 min read
Why an AI agent should be treated as another contributor to a governed platform, and how design-system mechanisms map one for one onto agentic equivalents.
Success Was the Incident
June 2026 · 10 min read
Adoption turned a five-minute browser cache into a budget problem. Splitting it into three caches with three owners, hashing dependencies behind stable component files, and a finops partnership cut request volume to a third of its peak.
From Snippets to Shadow DOM
June 2026 · 13 min read
Why a jQuery-era design system splintered into forks and customized local versions, how Shadow DOM moved governance into the component boundary, and what it took to earn adoption from a thousand engineers.
When the Model Is a Draft, Not the Source of Truth
June 2026 · 8 min read
Where AI output should stop in an audit-grade workflow: schema contracts at the boundary, override invariants in the database, and a review UI that treats model drafts as drafts.
Teaching AI Agents to Use a Design System
May 2026 · 9 min read
A practical look at constraining AI-generated frontend code with design-system primitives, agent instructions, and repeatable validation.
Compare notes
Working through a similar problem in system design, platform engineering, or applied AI? I’m always looking to compare notes on approaches, tradeoffs, and what worked in practice.