People keep asking me what I want to do next, and for the last two years I’ve been giving a rotating set of partial answers. I wanted to write down the clearest version I’ve managed so far, partly because saying it publicly might help me figure out the parts that are still wrong.

Up front: I’ve had a genuinely hard time defining this for myself. The question wants a clean noun back, product or operations or strategy or investing, and none of those words on their own have matched what I’m actually good at or what I want to keep doing. For a while I thought that meant I didn’t know. What I think was actually going on is that the shape of the role I wanted didn’t have a standard name.

My current best approximation, which is the one I’m willing to say out loud: I’m at my best when someone hands me one thing and lets me really take it. A product line. A launch. A problem that’s been sitting in the corner of a company for a while, getting worse. I pick it up, I figure out what it actually needs, and I drag it through until it works. What that’s been called has varied company to company. The underlying shape has been the same.

The closest external name I’ve found for it is Keith Rabois’s.

A bullet in that framing is a fast, iterative shot. A cannon is a slower, more deliberate, more consequential one. Rabois’s claim is that companies routinely hire bullets for roles that actually need cannons. They need one person to really take a thing, and they staff it with volume and process instead. I’m not sure that’s universally true. I do know that reading it was the first time I had a short phrase for the kind of role I keep wanting to be in, and for the failure mode I’ve lived through, which is being a cannon stuck doing bullet work.

When I look back at the work I’ve been proudest of, growing Fliff from 195K to 800K MAU, launching the first regulated sports stock market at Mojo, the Cerberus exit, the common thread isn’t the function or the seniority. It’s that I was handed one big, underdefined thing and trusted to own it end to end. In the roles where that wasn’t the setup, I wasn’t as good and I didn’t enjoy it as much.

Marc Andreessen has a related idea I keep coming back to, which is that for new things, the person doing it for the first time is often better than the person who’s supposedly already done it. The veteran is running a playbook from a market that’s already moved. The first-timer doesn’t know what’s supposed to work, so they have to look at what’s actually there. That’s part of why the domain I care about, AI and compute economics and the companies being built on top, is where I want to put the next few years. It’s new enough that nobody has a fully formed playbook yet.

Concretely, then: I want to be handed one consequential thing at a company I believe in, in that domain, and be given real room to run it. That could be founding product, or chief of staff with actual scope, or a first hire, or a role that doesn’t fit any of those cleanly. The title matters less to me than whether I get to own the thing.

I should say the obvious thing, which is that I’m still working this out. What’s on this page is my clearest articulation to date, not my final one. Some of it will sound different to me in six months. I’d rather publish it and keep revising than keep it in drafts. If the shape I’ve described resembles something you’re trying to staff, I’d like to talk, less as a pitch and more as two people trying to figure out whether it actually fits.