Blog

Writing from the team.

What we've been working on, what we've been thinking about, and the lessons that came out of either.

Shipping LLM apps without shipping LLM bugs

Evals, fallbacks, and tracing — three habits that separate teams who ship LLM features confidently from teams who don't.

Sync engines have come a long way. Here's how we choose between rolling our own and adopting one.

What's improved, what's still rough, and the patterns that have held up across three shipping products.

A design system is only useful if engineering builds with it. Here's how we make ours stick.

We tracked the engineering cost of every microservice we built across three years. The number surprised us.

Most engagements fail in the first week. Here's the scoping process we use to make sure ours don't.