Founding Lean

Even two years ago, building what we're building now would have required raising capital first. Not because the idea was more ambitious, but because you used to need to hire to fill the gaps in your expertise just to find out whether the thing was worth making. A front-end engineer. A backend engineer. Someone with enough domain understanding to make sound product decisions. You had to assemble a team before you could even test whether the idea was worth pursuing.

That's changed. Not because AI has replaced those roles, but because the friction from idea to working software has dropped so far that one or two people can move at a speed that used to require five or six. The work hasn't gotten simpler. But the cost of trying something, watching it fail, and trying again has fallen enough that you can stay in motion without burning through capital.

Building together

Earlier this year, I left my role at Sano Genetics so I could give this venture my full attention. Now my wife Bronwyn and I are building Flow Myna together. Bronwyn brings the domain expertise... years in event-based simulation, consulting, and process mining. A deep understanding of where organisations struggle when they try to figure out how work really happens inside their systems.

I handle the technical side: architecture, implementation, and the ongoing work of using AI tools well without letting them turn the codebase into something unmaintainable. I'm a heavy Cursor user; it lets me move quickly, but only when I'm intentional about where to let it help and where to take back control. Claude Code and others are always tempting, but exploring them properly takes more bandwidth than I'm willing to use while we're still shaping the core.

So what's this newsletter about?

This newsletter is documentation of this early phase and the ones that (hopefully) follow. Most people don't write during the uncertain part, when you don't yet know what will stay and what will get thrown out. But that's where the useful details are and I'm going to attempt to keep a log of what's working for me while I try to optimise my own workflows.

I'll write about the specific stack I'm using:

• Backend: Python and FastAPI
• Frontend: TypeScript and React

Some of it might translate to your setup. Some might not. Take what seems useful and leave the rest.

And I'll write about the judgment calls: where to let AI tools run, where to take back control, how to structure code so it doesn't turn into a monster. Because if you let AI make all the architectural decisions, then at least right now, you'll end up with a tangled mess that even experienced engineers would rather rebuild from scratch.

The productivity shift

I'm far more productive with these tools than I was without them. But they're evolving quickly, and my understanding of how to use them well is constantly changing. So this isn't me teaching. It's me sharing what I'm learning, and hoping you will share your tricks in return even if it's just a link to another related article that you found useful.

If you're building something similar, I hope this is useful. And please share your own approaches in the comments.