At a time when innumerable cryptographic projects are turning awkwardly towards artificial intelligence, Eric Winer has spent more than a decade building bridges between avant -garde technology and everyday users. Now, as cto close to AI, it is pursuing a vision where agents of the negotiation, communicate and make transactions in our name, a vision that could fundamentally remodel how we interact with the digital world.
Winer talks to the measured enthusiasm of someone who has arrived repeatedly to technological revolutions. After joining the Winklevos twins as Gemini’s first engineer in 2013 and helped to launch one of the first NFT markets before the term “NFT” became a common place, now confront what can be its most ambitious challenge: create a decentralized ecosystem of AI agents that preserve the user’s sovereignty in a future increasingly mediated.
Eric Winer is a speaker at consensus 2025. This interview has been condensed and slightly edited for clarity.
COINDESK: You have been in Crypto for more than a decade. How did that trip take you to focus on AI?
Eric Winer: I have been in the cryptographic world for 11 years. Eleven years ago I met the Winklevos twins through mutual acquaintances. They hired me as the first engineer in Gemini, who was the first exchange of Bitcoins regulated in the United States at that time. Our CEO mission was to take this fantastic blockchain technology (really only Bitcoin at that time) and apply some adult engineering around them so that institutions, governments and people can adopt effectively.
In 2013-2014, Coinbase was present, but it was a kind of Startup Rinky-Dink Silicon Valley in a moment before there really was a great Fintech culture even in California. Then it was not particularly built well. We were building Gemini in New York City with people who knew how money worked and trying to build not only for cryptography enthusiasts, but also for masses.
I have always been trying to obtain blockchain technology on the adoption barrier to get to a place where people would use it, perhaps under the hood, so that the world is more efficient and provides the most sovereignty people, even if they do not know.
COINDESK: And that took you to NFT before most people had heard the term, right?
Eric Winer: Within Gemini, we acquired a company called Nifty Gateway, which was basically the third Startup NFT after Opensa and Cryptokitties. We launched what was the first NFT market aimed at “Normies” in March 2020. It ended up growing in approximately 1000% month after the month for a few months, going from selling a thousand dollars of NFT to sell one hundred million dollars of NFT in one day.
This exposed many people to the world of collectibles and art online, just in the middle of the pandemic. From there, the NFT error bit himself and then because of the cryptographic error, taking them to the blockchain ecosystem as a whole.
Personally, I think NFT’s promise was unfortunately co -opted by the multitude of cryptography and speculation. Nifty Gateway never used the term “NFT” on its website or in its materials when it was launched. We fight that even becoming what was called. NFT is just one thing, but online. We were selling art, selling tickets. Having to assign a technical nickname and put it in public, I think, was actually counterproductive at the end of the day.
COINDESK: Many cryptographic projects seem to be turning awkwardly towards AI without a clear vision. What makes the close approach to AI different?
Eric Winer: Near Ia is a fascinating company because the founders of Near really started building an AI company. They started as an AI company, trying to train models to write code around 2017-2018. It was a bit early for that to be effective at that time.
Agi has always been the vision here. Near Ia is less a “crypt x ai” company than an AI company that is aligned with the vision with the same purpose closely. Near’s vision is that users have their own data, have their own decision making, are not in debt to corporate, financial or governmental interests, and can decide where their money is going and how their decisions are made.
Whether it is done through a block chain or is done through an AI model, both must have. If we do not have an artificial intelligence effort of user owned, then we will lose the game because everyone will start using chatgpt for everything in their world, and what Chatgpt uses for payments is whatever use for payments. Then, Blockchain loses because you did not earn the battle of ia, not the battle Crypto X Ai, but the true battle of the real.
Near Ia has interactions with blockchains, even close, because we believe that the chains of nearby blocks and another are a good way to do things. But we are mainly an AI company that sees cryptography as a technology; We are not a cryptographic company that tries to put AI.
COINDESK: What exactly do you offer as a platform at this time?
Eric Winer: Near Ia as it exists at this time, it is mainly an agents accommodation platform. Actually, it is one of the easiest ways to make an AI agent run online. Upload a python code or quite simple TypeScript code, and boom, it is hosted in line with a user interface, embeddable on any website.
It can be backed by any open weight model, although in general, we delegate the operation of the real model to other suppliers such as artificial and anthropic fires that are out there. The special sauce of our frame is that it facilitates that those agents of AI talk to other AI agents, either on our platform or out of place.
The other thing that is not yet live, but that will be when the consensus arrives, is that we are strongly invested in tees (trust execution environments). We believe that it is very important that you know where your data is going, that they are not harvested by these models for training, that even we as operators of a cloud system cannot look at their data, analyze or steal them if we wanted to. So we are putting all those agents and all the underlying models in these reliable execution environments so you can use them with confidence.
COINDESK: His vision seems to focus on AI agents who communicate with each other. How do you think that future?
Eric Winer: What we are trying to build within AI close to AI is an ecosystem of AI agents who speak with each other. “Ai agent” is a very overloaded term at this point. Most of the time when you see something that talks about itself as AFF, it is really a set of guidelines and tools for an AI model to act in your name. As, “I have models call or Claude or GPT, and I will instruct it to publish on Twitter in my name.” To do that, you need the ability to access Twitter and some general guidelines on how to read and publish tweets.
But that is not how I necessarily think of the agents. I think of the agents mainly in terms of how they interact with the rest of the world in their name. My AI agent is publishing on Twitter, managing my calendar, looking at all my messages and emails, deciding which are spam, what to answer, what to arise, as should be an assistant in my life.
The future we imagine, which I think will inevitably happen, is a world where most of our Internet access and the rest of the world is mediated by our personal assistants of AI.
COINDESK: That is a remarkably different vision of most cryptographic projects. How to remodeled our financial systems?
Eric Winer: If he is my agent of speaking, for example, Amazon’s agent, and they will buy something and negotiate, they are more likely to rationally choose a crypto -based payment method and probably a method of resolution of disputes based on AI and Credit based on AI. We reserve the financial system not building a better and trying to convince people to use it, but building a better one and trying to convince Ais to use it that is easier for AIS and for the people who operate them.
I think that is the cryptography adoption game now: we build a world of AIS that talks with AIS in your name, and then we convince AIS that the block chain and decentralization are a better world for them and for us than we have.
COINDESK: So, in his opinion, could it be the catalyst that drives the adoption of blockchain, instead of the other way around?
Eric Winer: I think that in the future, and probably not far in the future, you will see that vision that everyone seems to want: the super application, the only place where you go to everything you do online really becomes an assistant of AI. He is giving him a content of content, he is taking measures in his name, he is spending money on his name.
And it’s not just doing it alone. It is not just about hitting the API and the navigation of websites, but is talking with other AI agents. He is also handling his interpersonal communications, from person to corporate or from person to government. That must be a decentralized ecosystem, like the Internet, it is, to some extent, a decentralized system, but perhaps we can make one better in addition to blockchain -based records.
