Writing

Notes from
the pipeline

Engineering deep-dives, integration guides, and what we learned building an open document intelligence platform.

Why we open-sourced the entire document intelligence stack

Every parser, every extraction pipeline, and the MCP server, under MIT. What that decision cost us, and what it bought.

A frontier model reads a bank statement about as well as a careful intern. Here is where that breaks, and what we built instead.

Exposing 210 endpoints as agent tools does not work. What we learned designing a tool surface an agent can actually reason about.