We're starting a blog. Not because the world is short on AI content, but because the teams we talk to every day keep running into the same problems — and the practical answers rarely show up in vendor decks or research papers.
1. What you'll find here
This blog will stay close to the work:
- Evaluation patterns that actually catch regressions before users do.
- Field notes from PMs and engineers shipping LLM features in production.
- Trade-offs we've made building Plumloom — what worked, what didn't, why.
No hype. No "the future of AI" think-pieces. Just things we'd want to read if we were in your seat.
2. Who it's for
If you're a product manager, engineer, or founder shipping anything powered by an LLM, this is for you. Whether you're wiring up your first eval or trying to make sense of why your model regressed after a prompt change, we want this to be useful.
3. What's next
We'll publish when we have something worth saying — quality over cadence. If there's a topic you'd like us to cover, tell us. We read everything.
Thanks for being here.