Posts
All the articles I've posted.
-
Your AI Feature Is a Gamble Until You Can Measure It
An LLM feature is only production-ready when you can measure its quality, trace its behavior, and roll back a bad release. Here is why the demo proves nothing, the metrics that actually matter, and the roadmap for shipping AI features with engineering discipline.
-
The 2026 CSS You Can Finally Use
@scope, container style queries, :open, contrast-color() - a batch of CSS features crossed into Baseline this year and are safe for real projects now. Here is what each one kills, a practical example of each, and how to judge adoption timing yourself instead of guessing.
-
Interviewing Developers When AI Writes the Code
The whiteboard algorithm tests a skill the job no longer centers on, and take-home tests can no longer tell you who did the work. Here is what to assess instead - code review, judgment, and how a candidate actually uses AI - plus a practical interview loop for engineering managers.
-
Write Your CI Once - Reusable Workflows as DRY for Pipelines
CI configuration is the most copy-pasted code in most organizations - ten repos, ten near-identical pipelines, each drifting on its own. Here is how reusable workflows fix that, how to version and pin them safely, and the security payoff that makes it worth doing properly.
-
Design Tokens Are a Contract Too
Design tokens are the API between design and development - and like every API, they need a contract. Here is the three-layer token architecture, why the semantic layer absorbs redesigns and dark mode for free, and the governance that stops your token set from quietly rotting.