If A and B, then C
The first time I saw an algorithmic trading desk up close was in September 2000, at a hedge fund office in Westport, Connecticut. Programmers sat where traders used to, and execution algos sliced up orders automatically. The head trader told me something I've never forgotten: because of the algos, they had taken out half the trading floor and put in a half-court basketball court.
It felt futuristic at the time, but the market structure underneath hadn't moved. The algos still ran inside the same exchange hours, the same custody model, the same clearing cycles and the same settlement delays. Machines were optimizing inside rails designed for humans.
That memory came back to me this month, in a conversation at a conference in London. The person I was talking to agreed with Part I: AI will probably become the dominant user of stablecoins, and to her that felt intuitive. She pushed back when I took the thesis a step further. If stablecoins pair with tokenized securities, I said, then AI becomes a dominant user of tokenized securities too. Algorithmic trading is nothing new, she told me. Hedge funds have been automating execution for decades.
Fair point, but I'm making a different argument. Hedge funds upgrading their algos is a story about one corner of the market. What interests me is the much larger universe of plain-vanilla asset allocation: balanced portfolios, ETF rebalancing, risk-managed pension allocations and wealth management at scale. When the settlement layer changes, that kind of optimization stops being a specialist activity and becomes part of how the whole system works.
The proof
Premise A: AI becomes the dominant user of stablecoins. Machines optimize continuously. Cut-off times and batch cycles mean nothing to them, so if stablecoins are programmable, instant and always available, machines will prefer them.
Premise B: stablecoins pair with tokenized securities as the new settlement base layer. This has already moved past theory.
Superstate has launched a Direct Issuance Program that lets SEC-registered public companies issue new shares directly onchain in exchange for stablecoins. Investors receive tokenized equity immediately, and shareholder records update in real time through an SEC-registered transfer agent. That is primary issuance, onchain.
Securitize is enabling fully onchain trading of natively digital public equities. During US market hours, trades match on best bid and offer; after hours, liquidity moves to automated market makers, so the market is designed to keep running around the clock.
Nasdaq has formally asked the SEC to let tokenized versions of listed stocks and ETFs trade on its existing order books, with the same ticker, the same CUSIP and the same shareholder rights. Investors could choose to settle those shares into a blockchain wallet instead of only into a brokerage account.
The Nasdaq filing matters most to me, because it puts tokenization inside the national market system instead of on a parallel crypto venue. Exchanges, transfer agents, issuers and regulators are all moving. When the base layer shifts like that, the properties of the system shift with it: settlement can be instant, records update in real time, and markets can stay open.
Markets that never close suit agents far better than they suit people. There are 168 hours in a week. US equity markets are open 6.5 hours a day, five days a week, which comes to 32.5 hours, roughly 19% of the week. Add pre-market and after-hours trading and you get to about 16 hours a weekday, or around 80 hours a week, still less than half. And extended hours aren't continuous markets: liquidity is thinner, spreads are wider, institutional participation is limited and settlement stays periodic. Stitching the US, European and Asian sessions together doesn't produce one continuous system either, because liquidity fragments and clearing and custody stay jurisdictional.
An AI agent runs all week. Once capital is programmable and settlement is continuous, markets that open for less than half the time start to look like a design choice, and design choices can be changed.
Conclusion C: if AI dominates stablecoin settlement, and stablecoins settle tokenized securities, then AI becomes a dominant user of tokenized securities. Excitement has nothing to do with it. The plumbing now supports it.
Beyond hedge funds
The pushback I got assumes this is a hedge fund story. Hedge funds have automated execution for a long time, but their systems optimize within legacy custody, clearing and exchange hours. What is changing now is the constraint itself. With primary issuance, transfer agents and exchange settlement moving onchain, and securities clearing against stablecoins, optimization is no longer bounded by market hours and batch settlement.
That reaches well past trading desks. Take a wealth manager running diversified portfolios for high-net-worth clients. If tokenized ETFs settle onchain, an AI agent can rebalance each portfolio continuously against the client's risk profile, responding to drawdowns, shifts in volatility, liquidity conditions and macro signals as they happen. Today a lot of that work gets done once a quarter.
For pension funds the implications are bigger still. If tokenized securities allow real-time rebalancing across asset classes at lower cost, risk-adjusted returns improve, and better returns bring in more assets. In asset management, gathering assets is everything. Saving on fees helps, but improving outcomes is what wins mandates.
The three strands, sharpened
In Part I, I split fintech into three groups. This argument sharpens the lines between them:
Offchain fintech is optimized for people using legacy settlement rails, and it inherits their latency.
Adaptive fintech can operate across both systems.
Onchain fintech is native to tokenized assets, stablecoin settlement and agent execution, and it inherits their programmability.
As public equities, ETFs and other securities move into tokenized form, the advantage moves toward systems built for continuous optimization.
The extended theorem
If AI becomes the dominant user of stablecoins, and stablecoins settle tokenized securities, then AI becomes a dominant user of tokenized securities. Crypto doesn't need to "win" for that to hold. The market structure has changed, and once the base layer shifts, everything built on top of it reorganizes around the new layer.
Where this leaves me
If the base layer of finance is moving onchain, and AI becomes the dominant allocator of capital on that layer, I think the opportunity sits in the connective tissue more than at the surface.
At Norio Ventures, I'm not trying to predict which institution moves first. I'm looking for the founders building what sits underneath the shift: orchestration layers, compliance-native rails, and the abstractions that let wealth managers, pension funds and fintech platforms plug into programmable settlement without rewriting their whole stack.
Adaptive fintech will survive the transition, and onchain-native fintech will shape what comes after it. If I've got the geometry right, the winners will be the teams building for capital that moves continuously, and that world is arriving faster than most people expect.
A companion thread
If you've been listening to MoneyNeverSleeps lately, you'll have heard this thinking running through recent episodes:
2026 Crypto Predictions with Alejandro Gutierrez, on what happens once infrastructure becomes inevitable.
Avoiding Founder Burnout with Ollie Walsh, on staying adaptable when the environment changes faster than expected.
Scaling Human Expertise with Joni Pirovich, on intelligence as leverage instead of labour.
The Next Money Rails with Sean Lee, on programmable money in a world of competing sovereign systems.
None of those episodes used the phrase "the geometry of fintech," but each one came at it from a different side. The show is now video-first on YouTube, with shorter conversations and more than 300 back-catalog episodes sorted into themed playlists.
Originally published on the MoneyNeverSleeps newsletter on 26 February 2026. It follows Part I.
