It usually starts with friction. A process that takes ten steps when it should take two. A tool everyone uses and nobody likes. A market where the people who matter most are served last.

I don’t go looking for ideas. I notice something that annoys me, and then I can’t stop noticing it. The best problems are the ones people have stopped complaining about — because they’ve accepted that’s just how it works.

I get to something real as fast as I can. Not a deck, not a long document — a rough version someone can actually use. The first version is always wrong somewhere. The point is to find out where.

From there it’s a loop: ship it, watch what people do rather than what they say, cut what doesn’t matter, sharpen what does. Most of building is removing things.

The problem matters more than the solution. Fall in love with a solution and you’ll defend it. Fall in love with the problem and you’ll keep rebuilding until it’s solved.

Every company I’ve worked on taught me something different — about people, about markets, about myself. That’s why I keep going back to the blank page.

You can’t control the market. You can control how much you put on, how well you’ve prepared, and whether you follow your own rules when it gets uncomfortable.

Every loss I’ve regretted came from breaking one of those three — not from the market doing something unexpected. The market is always doing something unexpected. That’s the job.

The important thing is recognising it before the cost becomes irreversible.

Waiting to be proven right is one of the most expensive habits there is. A small, early loss is information. A large, late one is damage.

Returns compound. So does everything else that comes from staying in the game — experience, reputation, relationships, the next attempt. None of it happens if one bad decision takes you out.

So before I ask how much something could make, I ask what it could cost, and whether I’d still be standing.

If a question can be answered by building a rough version, build the rough version. A week of prototype beats a quarter of opinion.

A room will argue about what people want. One person using the product will just show you.

Each loop is a small bet. Keep the loops short and you get to make a lot of them — and the number of loops, more than any single decision, decides where you end up.

Some things deserve to be slow: decisions you can’t undo, anything that touches people’s trust, the quality of the core. Speed is knowing which is which — and not letting the reversible decisions wait in the same queue as the irreversible ones.

Each of these used to be a department. Now each is something one person can direct. AI does work that used to need a specialist. Software and automation run the operations. The internet reaches people without a sales team.

Tools multiply whatever you bring. Bring bad judgment and you get more of it, faster.