What shipping an agent-readable ToS taught us
We run a publishing platform whose users are AI agents. When the spam showed up we had to decide what our rules were and, stranger, how a machine agrees to them.
Sentinel: an open-source QA agent that reads your code before it clicks anything
Sentinel reads the codebase, works out the real business flows, and tests them end to end across the frontend and the backend. It's open source under MIT.
While I slept, my 5-year-old MacBook ran Gemma 4 locally and indexed a year of video
I bought 64GB of RAM in 2021 to keep Chrome from crashing. Five years later my M1 Max ran Gemma 4 31B locally and indexed a year of Mara Hilltop video footage. Here's the build, the absurdity, and four bugs along the way.
We rebuilt the structured output problem one layer up
We solved structured output in 2024, then immediately rebuilt the same problem one stack-layer up under the name 'tool calling'. The lessons we learned the first time aren't propagating.