AI, crypto, and the things that keep improving

It feels like everyone is pivoting to AI. The technology is fascinating, the opportunities seem enormous, everyone is pouring billions into AI startups, every startup is an AI startup that does this and that and makes millions in ARR, and apparently this also means that crypto is now dead and we should all move on.

Prices can certainly go down (I hope they don't though..), but I find it difficult to look at what blockchains can actually do and conclude that the technology is going backwards.

I have worked on infrastructure that allowed cross-chain token trading, so my perspective comes from building some of these things and dealing with their problems. Not with handling nodes and consensus and concurrency and Rust. I don't know where prices will go, but I do think the underlying capabilities of the blockchains and crypto as a whole keep and will keep improving, even when nobody seems particularly interested in talking about them.

Think about the simple ability to execute a computation, which can include moving money, and have a (globally) distributed network verify that it followed the agreed rules without one institution controlling the entire ledger. How cool is that? Add cryptographic proofs (zk and whatnot) that can verify certain things without exposing all the underlying information, and, well, that is pretty fascinating. Blockchains didn't invent distributed systems or cryptography, but they have pushed them into environments where strangers have real money at stake and every mistake is an invitation for someone to exploit it. This industry is the SpaceX of taking distributed computing and cryptography to the stars.

There is something close to what Nassim Nicholas Taleb calls antifragility here. Hacks, exploits, scams, broken designs, and people constantly trying to find the next weakness. Some projects disappear, and the people who lose money don't magically benefit from the lesson. But the teams that survive and actually fix what failed usually come out with better systems, and other teams can learn from those failures too. That is the part I find most interesting, and it can continue whether governments support the industry or make its life difficult.

Stablecoins were always the most obvious use case to me. If you spend some time understanding how money moves between people, businesses, banks and countries, you start wondering how something that looks so simple on a screen can involve so much effort underneath (and it's not always so simple on the screen either).

Cards, PayPal and banking apps made a lot of payments feel easy. But try moving money across the wrong combination of countries, currencies and institutions, and suddenly there are delays, fees, forms and someone asking you to wait a few business days. The experience depends quite a lot on where you happen to live.

A shared ledger that lets people transfer a digital representation of money without every institution maintaining and reconciling its own separate version seems like a fundamental improvement, one worth the fight that the industry has been doing all these years. Stablecoins still depend on issuers, reserves and somewhere to convert them into money you can actually spend, but moving them between wallets can be fast, cheap and available (and compliant) whenever you need it. And if dollar stablecoin issuers hold US government debt as reserves, wider adoption could also create more demand for that debt, which should give politicians their own reasons to find the technology interesting.

The part I never particularly liked was meme trading, especially the whole pump.fun thing. Even while working on trading infrastructure, I kept wondering why people would put so much money into something whose main selling point was that someone else might buy it for more.

I think I understand it better now. Humans have always looked for an accessible way to get rich quickly, and an app and a wallet make it pretty damn accessible. People try to make lot of money quickly with a meme that should go up, get disappointed, disappear for a few weeks or months, and then come back when the next interesting meme comes or discussion along. I don't have to like it to recognize the behaviour, and I don't think changing the technology underneath will make it disappear. It's just human nature.

What interests me more is what happens when AI systems start using the blockchains.

An AI that DOES things will eventually need ways to pay for some of them. If it can organize my day, why wouldn't it also order lunch, book a ride, or pay for a service it needs to finish a task? You know, the random things that we do during the day. Assuming those services accept the payment, I would rather give it access to a wallet with clear limits than let it move freely around my finances with just giving it my card number.

For example, spend up to this amount, interact with these approved contracts, and ask me before doing anything outside those rules. Blockchains with programmable accounts or smart contracts already provide building blocks for enforcing such restrictions. The model can make a bad decision, but it should not be able to talk its way around a correctly enforced spending limit. It's all about programmable constrains.

That is where I think advances in AI will be the next major play for crypto. Fast transfers, programmable permissions, and a ledger that software can interact with directly seem useful for systems that are supposed to act autonomously. Traditional payment systems can serve AI too (although with a significant friction), but blockchains offer another way to build this, with some interesting properties (ex. nothing offers 400ms transaction resolution like certain blockchains currently do).

Whether that makes a particular token worth more depends on what gets used and how the token benefits from that usage. Still, I think it gives the industry a much more interesting future than another round of trading memes (and there will be another similar round).

Like the memory (and general I/O) bottleneck, I think there will be an I/O bottleneck regarding transactions.

Thanks for reading!