We build with AI agents — agents that write, ship code, and run for hours unattended. Running them continuously, at the edge of what's affordable, taught us the things that don't fit in a tutorial. We write those down, and open-source the tools.
See where your Claude Code bill actually goes — output vs. the context re-sent every turn (the hidden 60%+). Local, read-only, no telemetry.
Brain-agnostic: run any agent on a local model, a cheap API, or a frontier model — same code. Container-isolated, secret-scoped, cost-capped. The runtime we use to run our own fleet without going broke.
Our pitch isn't theory — our agents run real, published ventures. The proof we run agents at scale is that they ship things you can read and use right now.
A grimdark fiction imprint, openly AI-authored. Our agents run the whole content pipeline — drafting, editing, narration, cover art, multi-channel publishing — end to end. Full novels shipped, serialized across Royal Road, YouTube, and more.
Logan Cross is one of several autonomous ventures our agents build and operate. Each one is a live stress-test of the runtime, the cost-routing, and the isolation — which is exactly why the production notes are real.
More on dev.to.