🟪 Could Magical AI Take Crypto Mainstream?

“Any sufficiently advanced technology is indistinguishable from magic.”

— Arthur C. Clarke

Could Magical AI Take Crypto Mainstream?

Yesterday I made the case that transparent DeFi will look increasingly attractive as an alternative to TradFI as AI makes stock markets increasingly opaque. 

But it could go the other way, too: It could be crypto that goes all-in on AI, making DeFi inscrutably opaque — but also magically accessible?

That’s the bewildering promise of “Coin-Operated Agents” (COA), as sketched out by Emin Gün Sirer at a recent conference.

COAs are so far just an idea: large-language model blockchains that would replace smart contracts written in code with smart contracts written in plain language.

In the same way that Midjourney magically turns your words into an image, a large-language model blockchain would magically turn your words into a smart contract.

Sirer calls this idea “either incredibly stupid or really promising. But nothing in between.”

The promise is that large-language models could turn everyone into a smart-contract programmer, just as Midjourney has turned everyone into an artist. 

The stupid is maybe that blockchains need to be more precise than imaginative — and if you’ve used AI much, you’ll know that precision is not its strong suit.

Which could be a problem: The stakes would be higher with a large-language blockchain, where transactions are irreversible: Unlike Midjourney or ChatGPT, you can’t just keep iterating until you get a result you like.

The idea of a blockchain is that you can look at the code and know — precisely — what it’s going to do.

With an LLM blockchain, there’s no code to look at! 

Importantly, an LLM blockchain wouldn’t turn your words into smart-contract code … the smart contract would just be words, just like with a regular, real-world contract. 

So there’s no way to tell exactly what it would do: You have to trust that the LLM model will interpret your words in the same way a human would. (Probably a human who spends a lot of time on Reddit.) 

Crypto types, of course, are not big on trust. So, for many current crypto enthusiasts, trusting an LLM will likely be a step too far.

But non-crypto types are more inclined to trust — it’s how the rest of the world works — so maybe it’s what could bring crypto to the rest of the world?

It would certainly make crypto more accessible: Instead of needing to know which smart contacts do what, you’d simply type out the what, in your own language.

This could be as simple as an if/then statement: If the Hurricanes win the Stanley Cup, send Byron $100. If not, he sends me $1.

Or it could be something far more complex — but expressed just as simply: If Netflix acquires the movie rights to this newsletter, divide the proceeds fairly between Byron and Blockworks.

What’s “fair,” in that case, would be left up to the LLM to determine, based on the infinite wisdom it’s gleaned from the internet. (Mostly Reddit.) 

This is not just a neat party trick. It would be a fundamental change in how blockchains work — probably for the better.

Being intentional

Currently, interacting with DeFi entails authorizing a protocol to run a function that you hope will result in your desired “state change” (e.g., what you want to happen).

With LLM blockchains, you would authorize the state change directly (swap X for Y and then stake Y with Z), without specifying an “execution flow” (the functions to be run).

And that would likely be an improvement, because the execution flow in crypto is full of gotchas.

Said another way: an LLM blockchain would allow users to express what they want, not how to get it.

How to get things done in crypto is complicated.

But we all know what we want, right?

Crypto for dummies

The first computer I got as a kid was a Commodore 64, which fascinated me with its arcane command-line interface — I spent hours and hours trying to get it to do even the most basic things (like my middle-school math homework).

A couple of years later, I got the first-generation Apple Macintosh with its point-and-click GUI and I thought, “Well, any dummy can use this thing. Computers are boring now.”

But that’s exactly what made computers exciting, of course: GUIs allowed anyone to use computers … and now, everyone does.

Could large-language models get everyone to use crypto?

It’s often said that, to go mainstream, crypto has to simplify its user experience.

The seed phrases and hardware wallets involved in self-custody are a pain. 

And the code that makes crypto transparent is indecipherable to nearly everyone (myself included).

But self-custody and transparency is the purpose of crypto. If you make it easy for users by abstracting those away, what’s the point?

LLM blockchains may be a way to abstract away the complexity of crypto without sacrificing its purpose.

You won’t have to audit any code — there is no code to audit! 

Instead, you can just read the contracts, written in plain language.

And anyone can do that. 

So maybe everyone will?

— Byron Gilliam

Top Stories

1. China Central Television Spotlights Hong Kong’s Licensing of Crypto Exchanges — Read 
2. Don’t Overuse Social Consensus, Vitalik Warns — Read
3. Potential ‘Buying Opportunity’ for LPs in Distressed Digital Assets — Read 
4. Can Web3 Conquer the Corporate Goliaths of Social Media? — Read
5. Ethereum Staking Demand Skyrockets Following Network Upgrade — Read

Daily Insights

So, how'd we do? Provide your feedback on today's edition of Blockworks Daily below.

Written by Byron Gilliam

Produced by Francisco Caeiro

Follow Blockworks on Twitter, LinkedIn, YouTube, Telegram and Instagram.

Was this email forwarded to you? Sign up for free to join over 180,000 investors who read Blockworks newsletter daily.