// Articles
Guides, tutorials, and comparisons to help you build better backend applications.
What a backend team's week looks like once agents write most of the code: specifying instead of typing, running the backlog in parallel, reviewing by running, and a platform team that sets rules instead of approving changes.
How deterministic provisioning, central guardrails, and a human at the merge let agents ship to your cloud without improvising infrastructure.
Why mocks let backend bugs through, the options for giving agents a working database and queues, and why isolation matters as much as the infrastructure itself.
How the two relate, where CI/CD strains once agents are producing changes, and why a factory uses your pipeline rather than replacing it.
The components of an AI software factory and how a change flows through them, from intake to a deploy on your own cloud, with the control plane kept separate from the loop agents work in.
No single product is a complete AI software factory yet. The layers one is built from, the tools that cover each, and how to decide where to start.
How to put the stages around your coding agent so a task goes from a request to a change deployed on your own cloud, starting from a service you already run.
The system that takes a task from an AI coding agent and carries it all the way to software running on your own cloud. What it is, one feature built through it, and what it costs.
Measured token spend across Encore, Express, Fastify, Hono, and NestJS when Claude Code builds the same realistic backend project.
Encore came in at $6.29 of token spend across three benchmark runs. NestJS came in at $12.69 for the same work with fewer passing tests.
We benchmarked Claude Code on five TypeScript backends. Cursor uses the same models, so the framework ranking and the numbers behind it port over directly.
We benchmarked Claude Code on five TypeScript backend frameworks. Only one shipped production-ready code on the first pass, and only one had a CLAUDE.md and MCP server in the box.