I have a confession, but it's not the one you're expecting. I haven't started six businesses this year. I've started zero. And working out why took me longer than it should have.
A friend asked what my "startup" was, at a dinner where everyone seemed to have one, and I genuinely didn't know how to answer. Not because I haven't built anything — I've built rather a lot, actually. I just haven't started a company for any of it, and calling it that felt like lying by overclaiming.
Three things that aren't businesses
There's Unload, a to-do app I built because I hated every other to-do app enough to make my own. Turns out that's rarer than it sounds — most people who build their own task manager stop using it within a fortnight. I'm still on mine, months later.
There's Match Wall, which started as a rugby stats page nobody asked for and is now a live, spoiler-safe dashboard you can open for any match, any sport, any fixture — the full stat wall, score blacked out until you're ready to see it. A proper little product, built across evenings.
And there's Sven, an AI health coach that lives quietly in my family's pockets and occasionally in other people's, entirely unmarketed, doing whatever it is unmarketed things do when nobody's watching.
None of these are companies. I don't have a pitch deck for any of them. What I do have is something odder.
I asked my own AI to mark my homework
One of my AI agents — I call him SHERLOCK, and his whole job is doing the due diligence I'm too attached to my own ideas to do properly — recently ran a proper audit across everything I've built this year. Real market comps, real numbers, no flattery whatsoever.
The results were not entirely flattering.
Unload is up against Todoist, which does roughly $26.5 million a year and owns the category outright with something like 300,000 paying customers. My honest ceiling, doing everything right, is hundreds to low-thousands of pounds a month — not an exit, and SHERLOCK didn't pretend otherwise.
Match Wall's underlying idea is proven at a scale I'll never touch myself — Twitter/X carried 147 billion impressions during the 2022 World Cup, 24,400 tweets a second on the France goal alone. The honest question isn't whether people want this. It's whether they'll leave Twitter to get it from me instead.
And Sven sits in a category with a genuinely enormous ceiling — one competitor, Simple, hit $100 million in revenue in 2024, up 64% on the year before. But SHERLOCK's actual verdict on Sven's own traction, in his own words, was that nine inbound messages a month with zero marketing behind them is "closer to noise than validation."
I paid an AI agent to tell me that one of my ideas is basically silence. Which is either the most useful thing I've built all year or a genuinely strange way to spend a Tuesday.
The word I was missing
That word — "validation" — is doing a lot of work in that sentence, and it's the word that's been bothering me all year. Because none of these are validated. None of them are businesses. But they're not nothing either. They're real, working, used-by-actual-people things sitting in the gap between an idea and a company, and until recently I didn't have a word for that gap.
I've started calling them draft businesses.
People like me — the ones who now have an AI workforce doing the actual building — want to test far more ideas than we used to, because for the first time testing doesn't cost anywhere near what building used to cost. But the vocabulary hasn't caught up. "Startup" implies you've already decided to commit years to it. "Side project" implies you've already decided it's a hobby. Neither one lets you just find out.
A draft business doesn't ask you to decide anything yet. It only asks whether it's worth becoming a real one — and lets the market answer that, rather than my own conviction, which is the input I trust least in this entire process.
Three drafts down this year. SHERLOCK's still totting up how many are actually any good. I'm not sure I want to know before Christmas.