Most advice about building a company is the answer to a puzzle. The hardest parts were never puzzles.
In 1921, the economist Frank Knight drew a line that still holds up. On one side: risk being outcomes you can’t predict but can put a number on such as a dice roll, or an actuarial table. You don’t know what happens next, but you know the shape of what could. On the other side: uncertainty being situations where you can’t even write down the odds, because the world isn’t stable enough to have any.
Almost a century later, John Kay and Mervyn King gave that second category a sharper name: radical uncertainty. The kind you can’t model your way out of, because the model assumes a world that no longer exists by the time you’ve finished building it.
Puzzles have answers but mysteries don’t
There’s a distinction intelligence analysts use that I keep coming back to. A puzzle is a question more information will resolve such as how many troops are at the border, or what’s our churn rate. Get the missing data and the puzzle dissolves but a mystery is a question no amount of information settles, because it turns on judgment about things that haven’t happened yet. Will this market exist in two year? Is now the time to raise? Should we build the thing nobody’s asking for?
For most of tech history, company-building was sold as a puzzle: get enough experience, hire the people who’ve done it before, run the playbook. Every “how we scaled” deck carries the same implicit promise: there is a right answer, and we found it, and you can too!
The AI era turned the puzzle into a mystery
When I was scaling Weedmaps from 30 to 300+ engineers, the environment changed slowly enough that a good plan could outlive its assumptions. You could forecast and by the time we took ShopKeep through a $550M acquisition, the ground moved faster but a quarter was still a quarter.
Is that all gone? The tools, the cost curves, the competitive set, the very definition of what a small team can build… all of it now rewrites itself faster than any plan converges. You are not solving for a stable world, in fact you are building inside one that reorganizes underneath you while you work.
You don’t beat radical uncertainty with a bigger team
The old reflex was to reduce uncertainty by buying expertise. Hire the COO who’s seen this movie. Add the CFO, the head of GTM, the agencies. Surround the mystery with enough experienced people that it starts to feel like a puzzle again.
It doesn’t work, and it’s expensive in a way that never shows up on the cap table. Every person you add to “cover” an uncertainty adds cognitive overhead and non-product surface area: coordination, alignment, meetings about meetings. Your forecasts get more confident, but they do not get more correct. You’ve spent real money to feel less uncertain about something that is still, fundamentally, unknowable.
You beat it by lowering the cost of being wrong
Here’s the move I wish I’d understood ten years ago: under radical uncertainty, the edge isn’t prediction but it’s adaptability. You can’t reliably be right more often but you can make being wrong cheap enough that it stops being fatal.
That reframes the whole job from the jump. Stop optimizing the forecast and start lowering the cost, in time and capital and political capital, of changing your mind when the world tells you to.
This is the actual thesis underneath Aqen. It is not “Aqen predicts the right path” but Aqen collapses the cost of execution across every function so that changing paths doesn’t cost you a quarter and a reorg. When a pivot is cheap, radical uncertainty stops being a threat you brace against and starts being a place you can operate. I’m building it because it’s the tool I needed in every job I’ve had.
The hardest decisions are always mysteries
The calls that keep founders up at night are never the ones a spreadsheet settles like whether or not to take the offer, whether the architecture you’re about to commit to survives the next eighteen months, whether the person you trust is the wrong hire for the next stage. They sit exactly on Knight’s second side, and the honest answer is nobody knows which is precisely why they’re hard.
It’s also why I started TideHelm. An advisor you call before the hardest decisions isn’t there to hand you a forecast and they’re there for the moment the data runs out and judgment has to start to help you ask what is actually going on here, which is the only real question radical uncertainty leaves you with.
If you’re building right now
Three things, concretely: stop spending energy trying to predict a world that won’t hold still; that energy is a tax. Audit your switching costs and attack the biggest ones to achieve the goal is to make changing your mind cheap. Separate your puzzles from your mysteries on purpose: throw information at the puzzles until they dissolve, and bring judgment, not more data, to the mysteries.
Radical uncertainty isn’t the obstacle to building in 2026. The sooner you stop trying to forecast your way out of it, the sooner you can start working inside it.
Let’s build.