Blog

Insights from the forge

Field notes from building AI systems in production — written by the engineers, designers, and operators doing the work.

When to build a custom agent vs. use ChatGPT

Off-the-shelf LLMs are powerful, but they hit limits fast in production. Here\'s how to decide when to invest in a custom agent.

How we built a retrieval-augmented support agent that resolves 67% of tickets without human handoff.

Three patterns we see in every failed automation project — and the questions to ask before you start.

Retries, fallbacks, cost caps, and the architectural patterns that keep production AI from melting down at 3 AM.

A look at how our engineering team uses Cursor, Claude Code, and custom internal tools to ship 40% faster.

When does self-hosting an open-source LLM make sense? We ran the numbers across three real client scenarios.

One email per month. No spam, ever.