I was never great at math.
I stumbled through algebra, barely survived trigonometry, and in 1990 I found the beats and lyrics of Public Enemy's Fear of a Black Planet far more interesting than anything in a math textbook.
Then in 10th grade, geometry clicked. It was the first math class I ever aced.
I didn't suddenly love numbers. Geometry just wasn't about memorizing formulas. It was about logic: if A is true, and B follows from A, then C follows too. You accept it and move on.
That way of thinking stuck with me. It probably explains why I later found statistics easier than long division, and why I'm drawn to systems that show their structure once you look at them from the right angle.
That geometry lesson came back to me earlier this month, on a stage in Amsterdam.
An onstage moment
I was moderating a panel at the ABN AMRO + Techstars Future of Finance Accelerator Demo Day, straight after a keynote by David Birch. Dave said something that stopped me in my tracks:
"AI will become the dominant user of stablecoins."
It sounds provocative at first, maybe even theoretical. But the longer I sat with it, watching nine finely tuned fintech pitches back to back, the more inevitable it felt.
Earlier in the discussion I'd made a fairly standard point: stablecoins solve real problems in emerging markets, where you can't trust the banks, the government or the currency, but those problems don't really exist in developed markets.
Dave pushed back. He didn't finish the thought, and he didn't need to, because earlier in his talk he'd dropped a stat that stayed with me:
£292 billion sitting in UK current accounts earning nothing.
Technology doesn't explain that number, and neither does regulation. Trust does.
Trust problems in developed markets are harder to see
In emerging markets, trust failures are loud and visible: inflation, capital controls, bank collapses. In developed markets they're harder to spot, but they're there:
Money sitting idle because inertia is profitable for banks.
Insurance priced opaquely because customers don't switch often enough.
Financial products designed around friction instead of outcomes.
We put up with it because switching is hard, optimizing takes time, and the system counts on individuals not paying attention.
An AI agent won't put up with it. It has no loyalty to bank brands or legacy relationships, and "this is how it's always worked" means nothing to it.
Tell your AI agent to optimize your finances (savings, payments, insurance, subscriptions) and it will do exactly that, continuously. And it won't wait for the banks to open on Monday morning.
Why stablecoins matter more in developed markets than we think
This is where the geometry snapped into place for me.
If AI agents are going to move money around on your behalf, reallocating, rebalancing, settling and optimizing, they need rails that are programmable, instant, available 24/7, cheap and deterministic, and that can be permissioned through code instead of paperwork.
Fiat rails, SWIFT and batch settlement can't do that. Stablecoins can.
On stage, I said something that landed harder than I expected:
AI agents are the only ones in the room who treat crypto like a grown-up.
People argue about narratives. An agent only cares whether the payment settles. Once you see stablecoins as machine money instead of consumer money, a lot of things start to make sense.
Lex Sokolin recently called stablecoins "robot money." It's a great phrase, because robots don't need brands, branches or business hours. They need reliable, programmable settlement.
What this means for fintech
Sitting through those pitches in Amsterdam, I came to a conclusion that's still reshaping how I think about fintech:
If AI becomes the dominant user of stablecoins, every fintech product will need onchain settlement.
Crypto doesn't have to "win" for that to happen, and regulation doesn't have to collapse. Optimization gets you there on its own.
I think we're heading toward three kinds of fintech, and the difference between them will matter a lot:
Offchain fintech, with beautiful UX wrapped around slow, permissioned rails.
Adaptive fintech, with products designed to plug into onchain settlement when needed.
Onchain fintech, with systems built natively for programmable money, AI agents and continuous optimization.
Stripe sees this clearly. I read its acquisitions of Bridge, Privy and Metronome as an infrastructure strategy more than a crypto strategy: Stripe is building the plumbing that lets money behave like software.
For a fund like Norio Ventures, this doesn't mean chasing the next Stripe-sized outcome. It means backing the smaller, sharper teams building the connective tissue (the ramps, abstractions and orchestration layers) that lets institutions and fintechs move onchain without blowing themselves up.
If all of finance is moving onchain, fintech is going with it, whether or not we like the conclusion.
Where this leaves me
Most major financial innovations have been institution-led: credit cards, electronic trading, derivatives. Crypto flipped that and started with individuals, and AI is speeding the shift up again.
Customers will move first and machines will move faster. Institutions will meet them there, or end up as background infrastructure. If A, and B follows from A, then C.
A companion conversation
While I was working through all this, I recorded an episode of MoneyNeverSleeps with Garrett Cassidy.
We never called it "the geometry of fintech," but we came at it from several directions: stablecoins, AI agents, geopolitics, dollar dominance, and why programmable money changes who actually holds power in the system. It's a good companion if you want to hear the thinking unfold in real time.
Watch on YouTube, or listen on Spotify or Apple Podcasts.
Originally published on the MoneyNeverSleeps newsletter on 18 December 2025. Part II follows.
