110 Days

In 2023, my boss at MongoDB sent me an article about LLMs and hallucinations. I was running product for the MongoDB query engine and Atlas Search at the time. I read it twice with total fascination. Then I forwarded it to a colleague.

“That’s nice,” he said. “But is there money in that?”

The academic in me wanted to clear the calendar and just research.

I used to joke that my dream job was sitting beneath a tree with a pile of books and a notepad. Just reading, writing and thinking. The response to that was always polite.

This September I left a demanding role with an amazing team to find out.

A predictable paycheck traded for something undefined. At Microsoft, Elastic, MongoDB, the interesting problems came to my doorstep. The storm found me. Now I have to go find the storm.

110 days in, it isn’t the tree exactly. More like an independent research practice. Dimension cliffs, cross-encoder tradeoffs, agent workflows. Deep dives into problems…

I’m still wired for 12-hour days. That hasn’t changed. What changed is fewer meetings and deadlines, no chopped schedules, and time that’s actually mine to allocate.

That time cuts both ways. More space to think also means more space to doubt. When my calendar was full there was no room to wonder if I was on the right path. Now there’s room.

What I didn’t expect: the creativity came back. Projects that had nothing to do with work. They’d stopped coming somewhere along the way. Now they come with ease again.

The plan for 2026? More research and more writing, ideally with more courses too. Whatever interesting problems arise, in whatever format they arrive in.