If you're reading this, you've stumbled onto the first post on a blog I've been meaning to start for a while. The plan is simple: a place to think out loud about AI — the parts that actually matter once you move past the demo, and the parts I find genuinely interesting from the engineering seat.
What I'll write about
My day-to-day is in the messy middle of AI: turning capabilities that look magical in a notebook into systems an organization can quietly depend on. That's the lens you should expect from most of what shows up here. A few themes I keep coming back to:
- Agents in the real world. What actually works when you give an LLM tools, memory, and a goal — and what falls apart the moment it leaves the demo.
- Retrieval that doesn't lie. RAG is easy to stand up and hard to make trustworthy. I want to write about the boring parts that decide whether it ships.
- Evals and observability. The unglamorous infrastructure that separates "it worked once" from "it works on Monday morning."
- Enterprise reality. Governance, access controls, adoption, and the human side of rolling AI into an organization that wasn't designed for it.
What I won't do
I'm not going to chase every new model release or write breathless takes on whatever launched yesterday. The pace of the space is exhausting enough without another newsletter adding to it. If something genuinely changes how I work, I'll write about it. Otherwise I'd rather go deep on fewer ideas than skim across many.
A note on perspective
I'm writing as a practitioner, not an analyst. Expect concrete examples, opinionated takes, and the occasional admission that something I tried didn't pan out. If a post here saves you an afternoon of debugging or shifts how you think about a problem, it's done its job.
More soon. If something here resonates — or you disagree with it — the contact page is the best way to reach me.