I Will No Doubt Get Lots Of Productive Work Done

Bowl in the Form of a Llama with Geometric Motifs, c. 600-1000, south coast of Peru or northern Bolivia

So the reason for my busyness that I’ve been teasing the past two weeks is finally public (enough): after an intense couple weeks of interviewing, I’m leaving Descript, and I’ll be joining Vanta on April 20.

I have complicated feelings (that I’m happy to share if you hit reply…) but at the moment I’m just excited to get a few weeks off in between, where I will no doubt get lots of productive work done and not just play Slay the Spire II… right??

The rest of this newsletter is something I’ve been working on the past few days: a formal “AI policy” governing my use of LLMs and, more importantly, laying out a set of principles for how I think about it. It’s a fairly long doc, but it was a useful exercise to work through it all, as things stand in early 2026, and I plan to keep the evergreen version up to date. That said, the tl;dr is that I’m comfortable using agentic coding tools like Claude Code, and I use LLMs for some writing-adjacent tasks, but every word you read from me, here or elsewhere, comes directly from my brain. And after this, I promise I will not talk about LLMs again for at least a month 🙂

In the email version of this newsletter, I included the complete text of my AI policy inline. But you, dear archive reader, can go to the evergreen version.