If David Johnston believes 2025 will be a pivotal year for smart agents and decentralized AI, it’s a pretty safe bet. Johnston is an open-source contributor to Morpheus, a platform that launches smart agents. He co-founded Bit Angels to invest in Bitcoin startups in 2013 and coined the term dApp one year later.
Johnston was attracted by Bitcoin’s utility as an open-source, peer-to-peer system. He wasn’t the only one, as a community of innovators believed in an open-source, peer-to-peer blockchain with its own token that pays for the hardware while maintaining independence and censorship resistance.
“I think that the real genius of Satoshi was creating that way for the platform to pay its bills; it was a real breakthrough,” Johnston said.
Innovators soon learn that timing is as important as an idea. Johnston recalls an early attraction to AI in 2011, but it was too early; the technology couldn’t scale. That changed a decade later when ChatGPT arrived. In 2022, Johnston wrote a paper on smart agents, charting a path for a technology that will soon be pivotal to billions of users.
Smart agents explained
“It’s the idea of an agent that could interact with smart contracts, act on behalf of the user, and take economic actions by connecting the user’s Web3 wallet to their large language model,” Johnston said. “I worked with some developers, and we built the first smart agent that could use the large language model as a brain and a wallet to do transactions and act on behalf of users.”
A decentralized network is key, with a community soon rallying around the idea. Johnston said ShapeShift founder Erik Voorhees backed the idea. Such AI interfaces will help popularize decentralized money because the AI interfaces are the ideal way for the average person to talk to the smart contract via their wallet. No technical knowledge is required.
“This is how we get to a billion crypto users; you make it easy to use,” Johnston said. “You need that Netscape moment. You need that search engine moment, like the early days of the web, where it went from something for people very technical to something anybody could use.”
You need that Netscape moment. You need that search engine moment, like the early days of the web
“This feels like that. That’s the moment that we’re at now. But there’s no company, there’s no foundation, there’s not even a DAO. It’s just an entirely decentralized, open-source project that people have been contributing to and building for the last year.”
Timing is everything
Johnston said the timing was serendipitous. If some smart agent technology was absent, a solution soon came. Needs were met like this multiple times.
Vision and talent were, of course, even more important. The Morpheus paper described a clever mechanism whereby launch supporters could keep their ETH but direct the rewards to bootstrap the project while competing for tokens.
“It bootstrapped the whole network into existence,” Johnston said. “You didn’t need a token sale, didn’t need an entity, you didn’t need an ICO. All of a sudden, you had the building blocks to just do it all in the smart contracts and in a non-custodial way.”
Then came Bittensor and Hyperbolic. Such GPU networks have connected to Morpheus and provide AI models to interested users.
Private AI networks also have a role. Johnston said Voorhees developed Venice.ai. It’s similar to ChatGPT but private, with encrypted prompts travelling back and forth to the GPU. Data isn’t stored. Johnston said it’s personal AI using open-source models. Open-source models have caught up to the proprietary ones.
Tipping points
This is how smart agents, decentralized AI, and cryptocurrency will come together to help everyone, but especially grow utility in the developing world, where someone can’t afford $300 a year for ChatGPT.
“That’s where the usage is taking off,” Johnston said. “It’s really cool to see.”
Johnston said decentralized AI has several advantages. Like Linux and Android, it’s cheaper and better. Centralized AI systems are easily censored and siloed by governments. That’s why Bitcoin took off in Asia and Africa; the government couldn’t turn it off.
“A couple billion people aren’t going to be able to use centralized AI,” Johnston said. “The only thing they’re going to be able to use is decentralized AI.
“Even if the government’s not blocking (centralized AI), the cost is a barrier for a couple of billion people, the guy who can’t afford $200 a month. Sorry, guy in Africa, guy in South America, guy in Southeast Asia. Even if your government didn’t block it, you can’t afford it. We’re unlocking those four or five billion people who are in that category.”
Johnston sees the best outcomes where everyone has a personal smart agent that is working for them, not the government of a big company. Someone will control your data, so it might as well be you.
Compare it to everyone having, not renting, a smartphone. Johnston said the device becomes part of your brain; it’s loaded with photos, finances, social connections and work.
We’re getting to the point where smart agents are ubiquitous. It’s in the background delivering results. Like the smartphone, we don’t need to know how it works. It’s simple and productive.
Johnston believes a company like X will put smart agents over. The company is pursuing money transfer licenses, and when it adds a wallet, watch out.
“When they add a wallet to X and start enabling anybody to use this technology, it’s will force all the other tech companies (to step up). It’s 2024, and Google still doesn’t have a web wallet. It’s amazing to think, right?
“They can’t even use Web3 because all these tech companies haven’t wanted to touch money because it’s so regulated and onerous. But, one by one, the crypto industry is knocking those barriers down. Decentralized, self-custody solutions have gotten good enough that they’re not barriers anymore.”
The smart agent surge has begun
Humans have free will; they have goals. Describe them to a personal AI, which will try to fill in the gaps before delivering the results. Johnston sees this as the best way to organize it.
2024’s last few months have delivered a surge in AI agents. Johnson said they could start as simply as a meme on X. Someone connected a large-language model, and they got down to business.
“I want my smart agent to work like 100 of me,” Johnston said. “What would have taken me 100 hours to do it does in a minute?
“That’s the value here. Smart agents are an extension of your brain. It’s multiplying your brain power 100X, 1,000X. So very quickly, everybody’s going to want a smart agent because if you don’t have one, it’s going to be like not having a smartphone.”