Why I'm building in public
AI coding agents · updated may 2026
I’ve shipped things quietly for most of my career, and I only started writing about the process this year, after enough people asked how RaceLink actually got built that I realized I was repeating the same conversation privately instead of just writing it down once. That’s the boring, practical reason. There are a few less boring ones underneath it.
The first is that building in private lets you avoid deciding what you actually think. When nobody’s reading, you can leave a technical decision half-justified in your own head — “I’ll figure out why later” — and never circle back. Writing it down for an audience, even a small one, forces you to actually articulate the reasoning, and more than once that process has surfaced that my reasoning didn’t hold up. Public accountability isn’t really about other people watching. It’s about not letting yourself get away with hand-waving.
The second is that shipped-and-polished case studies are a genre I don’t trust anymore, including my own past ones. A finished write-up of a successful launch reads like it was inevitable, because you’re only telling the part of the story where the decisions worked out. It’s not dishonest exactly, it’s just missing the part where you tried the wrong thing first, or shipped something broken, or picked the wrong pricing model for four months before fixing it. That part is the actually useful part for anyone reading it who’s earlier in their own build than you are. I’d rather write the version with the wrong turns still in it.
The third is more selfish: I run four different things at once — RaceLink, HostLink, Liquid Lava, plus leading AI tooling at my day job — and writing regularly about what I’m building is one of the only mechanisms I have for making myself stop and actually evaluate whether a decision from three weeks ago was right, instead of just accumulating decisions and never revisiting any of them. A weekly build log is a forcing function for reflection I wouldn’t otherwise schedule.
None of this is a strategy document about audience-building or content marketing, and I’d be lying if I framed it that way. It’s closer to a discipline I adopted because the alternative — building everything quietly and only talking about it once it’s done — was letting me avoid thinking clearly about my own decisions. If it happens to be useful to someone else reading along, that’s a genuinely good outcome, but it’s not the reason I started.
builder, codelabs.com.au
Stay up to date with AI coding
New articles roughly every couple of weeks. No spam, unsubscribe any time.
Related