Blog
Building Cascade in the open
Engineering notes from building an open-source AI app builder that runs on your machine: what local models get wrong, what a real build looks like, and what we measured along the way.
Build log · 6 min read
One prompt, one turn, a working store: inside a real Cascade build
Plan, code, build, browser checks — every step of a twelve-minute run, from its own trace, and the bugs the earlier runs exposed.
Engineering · 7 min read
Small local models fail the same way in every harness. Cascade is built to carry what they drop.
The same 9B Ollama model through two coding-agent harnesses: identical failures, very different recovery. The measurements, and how to run agents locally.