Matheus Pagani Wants to Teach the World About AI and Web3

The founder and CEO of Venture Miner believes the combination of the two buzziest (and often misunderstood) technologies can be brought together to create useful products and services.

When I tell you to describe something in a way that allows people to get a clear picture, I mean really clear. Like, if I were to ask you to describe your favorite cookie, I want you to tell me about its sight, sound, smell, touch and taste. You might not realize it then, but describing something in this way is actually quite hard.

Let me try to describe Matheus Pagani, founder and CEO of Venture Miner. Matheus is a male with light caramel skin and dark brown hair. Even though his hair is cut close, you can tell it’s curly. He’s got a thick dark brown, almost black beard, which connects to a mustache. His eyes are dark brown behind thin wire glasses. His bottom lip sticks out a little further from his top lip, giving him a look of assurance, but not arrogance.

Picturing him yet? How confident are you?

Oh yeah, and he’s Brazilian.

Got it?

Let’s see what Matheus Pagani actually looks like.

Is this what you had come up with in your head from my description? Doubt it. Whenever I told you he was Brazilian, did you accessorize him in bright colors and a feathered headdress? Something like this?

If so, check your bias, but also you’re thinking like an AI. That was what ChatGPT came up with from the prompt “some Brazilians having fun.” Pagani showed this and other examples spit out by our generative AI (Italians have fun by sitting around long tables with multiple generations eating pizza) during the AI2Web3 Bootcamp in NYC in early December.

The bootcamp, run by Pagani and Build City, brought together 59 participants across all skill levels to learn how the two buzziest (and often misunderstood) technologies can be brought together to create useful products and services. Pagani used a version of the middle-school assignment to explain how and why AI made the significant leaps that have kept us all excited and on edge over the past few years. Before there was largely only text data being used to train AIs, and as the exercise highlights, that only goes so far. But mix text information with visual data, and you get a fuller picture.

And understanding this, getting hands on with both AI and blockchain technology to understand its core components is what the bootcamp was all about. For Pagani, these skills are going to be relevant for nearly all people – engineers, tech users, journalists, artists, doctors – real soon.

“We want to join brilliant minds from all backgrounds to come and work with AI and Web3, since the junction of their multiple perspectives can uncover new use cases that we would never envision just with a specialized Web3 or AI mindset alone,” Pagani said. “Nowadays we have tools to easily enable any non-technical enthusiasts to build practically functional applications and systems just with "plain English," so what matters is bringing passionate people interested in solving problems together with the proper education. When you have this combination, you just need to light the match and watch it burn.”

Mind-Boggling Building

What makes the intersection of these two technologies so exciting is just how much you can build in such a short amount of time without really any prior technical experience.

Not only will AI source whole codebases with the right prompt, but the crypto industry is also building tools to help make developing at the intersection of both more intuitive and accessible.

For instance, Coinbase, who sponsored the bootcamp, launched AgentKit in November. The framework allows developers to build AI agents with their own crypto wallets, enabling the agents to interact autonomously with blockchain networks. This could be used to build a squad of agents that can monitor the markets and execute trades automatically based on predefined rules and guardrails.

“One day, we’ll have AI agents own their own cars and operate their own taxi service that gets paid by customers in crypto and then uses that crypto to purchase repairs,” Lincoln Murr, associate product manager at Coinbase, told the attendees.

Coinbase currently has a grants program ongoing for building with AgentKit. “What you build doesn’t have to be useful; we have a bias towards cool stuff,” Murr told the bootcamp, hoping to inspire projects and applications that no one has yet thought of.

Ora Network also has an interesting model for developers looking to build AI-enabled Web3 applications or vice versa. The network allows developers to utilize current large language models, including Meta’s Llama3 and Stable Diffusion, but it also enables developers to build their own models and offer a so-called initial model offering (IMO) to crowdfund its continued development.

“It’s kind of winner-takes-all right now in AI, but with this model, we’re allowing the crowdfunding of AI building and training, so people can have a share of the models, which is empowering if we think these models will run society in a decade,” Alec James, partnerships and growth