Last month, I started full-time on a new startup. It’s early days, but we’re having a lot of fun.

A startup, fundamentally, is a search for a repeatable, scalable business model. You rapidly try things, run experiments, learn, and iterate your theories about how to build a useful product that people love.

This experimental nature makes many founders reluctant to talk about their startups early on. What if I share our initial ideas, but they turn out to be wrong? It can seem nothing is certain enough to talk about.

For Jenn and me, though, a few things do seem certain:

1. We want to build software that people love
2. This software needs to be valuable enough that it can support an incredible team
3. LLMs have created a horde of opportunities to build great, useful products
4. Sharing what we’re learning is worthwhile

That makes for a great initial sketch of where to direct our work.

Getting started

A delightful part about learning new things is that knowledge has a fractal shape. Recently, we started with a high-level curiosity:

How do you build great products on LLMs?

Researching this uncovered many deeper problems and questions, for example:

How do you test and evaluate a non-deterministic LLM-powered product as your team iterates?

Inside this lay countless finer-grained questions around evaluation approaches, workflows, safety, developer experience, testing, product experiment velocity, and so on.

In more-established domains, researching fractally into a problem uncovers layers of increasingly detailed answers and best practices. When you’re researching an emerging area, though, once you get more than one or two questions deep – even talking to world experts – you hit on meaningful unsolved problems.

In the interviews I’ve been doing, I often get answers like,

Oh, I don’t know! I would love to hear what other teams are doing for that.

or

I saw a paper on that, but I don’t know of anybody who has tried implementing it.

or

If you can solve that well, can we be your first customer?

Those are exciting things to hear!

So we’re spending our days writing code, learning, and talking to engineering teams who are building with LLMs. Worst case, we learn a lot about how to apply a useful new tool, and help some folks along the way. Ideal case, the product we’re prototyping helps those teams ship awesome products faster, kickstarting a great business.

If you’re a developer who wants to try what we’re working on when it’s ready, or compare notes on the challenges of applying LLMs on product teams, get in touch!

Or, you can just follow along as I write about the wild adventure of starting something new. 🚀