My founding thesis

I've caught myself saying the same five things over and over. They started as observations and hardened into a filter. Every idea I'm looking at right now goes through it. If something clears all five, I stop arguing with myself and start.

Here they are.

It has to survive a downturn

I started a company three weeks before COVID. I'm not doing that again.

Whatever I build should hold up when the economy gets worse, or ideally do better. Most markets don't have this property. Their demand is tied to optimism, headcount, discretionary budgets, the things that vanish first. A few markets get more important when money gets tight. Those are the ones worth a hard look. If I can't explain why a recession helps my idea, I treat that as a missing answer, not a detail.

It has to be non-obvious

The obvious ideas are already taken or already commoditized. By the time an idea is easy to agree with, the window is closed.

I want the thing that sounds slightly wrong today and obvious in a year. That gap is the opportunity. So I've started reading agreement as a warning sign: if a smart person nods the moment I describe an idea, I've probably missed my window. The ideas worth chasing earn a pause, maybe a flinch, before they earn the nod.

It has to have a moat

Cheap software with no defensibility is a race to the bottom, and I don't want to run it.

A real edge looks like a few specific things: technical depth that takes years to rebuild, regulation that's expensive to clear, distribution nobody can copy, or, best of all, a dataset nobody else can collect. I keep coming back to data. You can't simulate the real world. Whoever captures it honestly, and keeps capturing it, gets to keep the lead. Code gets cloned in a weekend. A proprietary stream of real-world data does not.

It has to help people, not replace them

I don't want to build something that quietly puts people out of work. I'd rather make someone better at their job than make their job disappear.

I'm holding a paradox openly here. The automation comes eventually, no matter who builds it. Pretending otherwise is naive. But there's a difference between the two paths to it. I'd rather the people gathering the data be the ones actually serving the worker first, earning the right to that data by being useful, instead of extracting it to replace them as fast as possible. Same destination, different relationship with the people who get you there.

Capital is fuel, not the point

I'm not anti-money. I'm cautious about what money does to a company.

Capital is fuel. It's worth taking when you need to move fast enough to lock a position in a market that only has room for one or two winners. That's a real reason. Outside of that, it mostly buys obligations: a board, a growth curve you now have to hit, a clock you didn't set. So my default is to bootstrap and stay free, and to raise only when the market itself forces the timeline.

The last part

If an idea clears all five, I stop trying to talk myself out of it.

That sounds minor. It isn't. The filter exists to kill weak ideas early, but a filter that good can quietly kill strong ones too, because there's always one more objection to find. The discipline isn't only in the five tests. It's in knowing when the testing is done and the building starts.