Key Takeaways
In 2023, Avalanche founder Emin Gun Sirer introduced the term “coin-operated agents” to describe on-chain AI agents that can be triggered or incentivized by crypto transactions.
While the technology remains experimental, in an interview with CCN, Gun Sirer expanded on how the concept has grown alongside advances in AI capabilities.
The premise of coin-operated agents on Avalanche is that specially-equipped blockchain validators will be capable of running Large Language Models (LLMs).
As Gun Sirer explained, at its heart, the concept is about enabling everyday users with little to no coding experience to program smart contracts—essentially vibecoding for blockchain.
“What coin-operated agents bring to the table is a blockchain system where the blockchain is programmed in a natural language,” he emphasized.
In Gun Sirer’s view, one of the things holding back wider blockchain adoption is that transactions that are easy to describe are often difficult to program.
Consider, for example, a basic agreement that X will pay Y a sum of money if they can first raise a predetermined amount elsewhere.
This is exactly the kind of deal that smart contracts should be able to facilitate. But the vast majority of businesses today still rely on multi-step, human actions to carry out such contract requirements.
Coin-operated agents could massively expand blockchain use cases and smart contract adoption.
However, developing the technology is “fraught with all kinds of issues,” Gun Sirer observed.
Creating a blockchain with LLMs built in at the infrastructural level throws up some unique challenges.
“What if AI starts to hallucinate? What if AI has other ideas? What if it decides to use the money for its own purposes?” These are some of the questions Ava Labs has been dealing with, Gun Sirer noted.
To realize Gun Sirer’s vision for coin-operated agents, Ava Labs is building a new Avalanche subnet that integrates LLMs at a fundamental level.
The ambitious project would expand Avalanche’s reach in the emerging decentralized AI space, where a few platforms like BitTensor currently dominate the conversation.
Although Avalanche isn’t necessarily the first blockchain protocol that comes to mind when thinking about decentralized AI, there are a growing number of subnets, including Kite AI, Loyalty+ and Turf.GG, that are specifically designed for machine learning tasks.
As Gun Sirer sees it, such initiatives are akin to developments happening across the digital universe, where blockchains are evolving from a niche solution to a widespread feature of the modern internet.
“What we’re seeing with blockchains is that they are the next step in the evolution of large, complex software systems,” he stated.
For AI applications and beyond, he predicted that client—and server-based models will increasingly be superseded by blockchain-based ones.
As this happens, the chains themselves will start to fade into the background as technologies like coin-operated agents abstract away the complexities currently associated with the technology.