How Dropbox is moving towards agentic systems that can execute scoped tasks, and how we’re building platforms to support those workflows.

About 10 years ago, I realized all the best programmers I had worked with had something in common: they were fast.

The MCPs that make Claude Code useful for iOS, and a working app on your phone by the end of the afternoon.

How to minimize custom prompting and rely on upstream tool updates.

Interactive config generator for Claude Code environment flags. Find your preferred balance of cost, tokens, and thinking.

Why predictable CSS state resolution matters, and how Tasty approaches it with state maps and non-overlapping selectors.

It still hits like a ton of bricks to see the steep decline in Stack Overflow questions. What does that mean about learning in our industry?

On coding agents, malleable software, and the future of interface invention.

Remaining vigilant about cognitive debt and atrophy.

The other month I finally ran an experiment we had been postponing for over a year at .txt.

What the PNR locator on your boarding pass actually contains, and why fare calculations are written in a currency that does not exist.

It doesn’t mean you can’t get AI to help with accessible code, you’ve just got to know what you’re doing.

Devs are purposefully burning tokens (and money!) to inflate their AI usage and hit AI usage metrics which they treat as targets.

Talking websites into existence.

AI coding tools are optimising the wrong thing and nobody wants to hear it.

Why we need collaborative AI engineering.

A collaboration model that shined a light on how things could (no, should) be.

Most of us are still running product teams on a model that was designed for a problem we don’t have.

Agent sessions work well for focused tasks, but most real projects are too broad and complex for a single context window.

Two practical approaches to the same insight: memory is a design problem, not a tooling problem.