AI agents fueled a frenzy of startup building at the Consensus Miami EasyA hackathon

MIAMI BEACH, Fla. — At the EasyA Hackathon tucked inside Consensus Miami 2026, the energy felt less like a traditional crypto developer event and more like a live audition for the next generation of the intersection of blockchain and AI-native startups.

Nearly 1,000 developers competed at the venue, some from established crypto ecosystems like Base and Solana, and others arriving from companies like Microsoft and Google, all racing to build products around one theme that kept surfacing in conversation after conversation: AI agents.

The focus on AI agents had already emerged earlier this year at the EasyA x Consensus Hong Kong hackathon, where organizers described 2026 as the “Year of the Application Layer” as developers increasingly shifted from infrastructure tools toward AI-powered consumer applications and autonomous agents.

For brothers Dom and Philip Kwok, co-founders of EasyA, that evolution is exactly the point. What began as a small hackathon series in Austin, Texas, during Consensus 2023 has quickly transformed into one of crypto’s most closely watched builder gatherings, attracting young passionate developers with increasingly teams with serious technical pedigrees.

Their ambition for the event is bluntly simple. “We want billion-dollar companies coming out of EasyA,” Dom Kwok said during an interview with CoinDesk at the hackathon floor. “We’ve already had, out of our other hackathons, a $10 billion company.”

That success story has become part of EasyA lore. One Harvard team that pitched at a previous EasyA event went on to found “Cognition AI,” which the Kwoks say is now valued at roughly $10 billion. Another former participant, Axal, is building stablecoin yield products backed by bitcoin.

Other alumni have reportedly gone through Y Combinator, raised from top venture firms and processed hundreds of millions in transactions. The message to developers walking through the Miami event was clear: this is no longer just a couple-days coding competition, it’s increasingly being framed as a launchpad for venture-scale companies.

This year, however, the center of gravity has unmistakably shifted toward agentic AI. Coinbase sponsored challenges around x402, an emerging framework developers are experimenting with for AI-agent payments and interactions, while Solana and Solana Mobile pushed teams toward mobile-first applications and consumer experiences.

“Lots of developers [are] really excited about AI agentic workloads,” Dom said, pointing to the recent wave of massive venture funding flowing into AI-agent infrastructure startups.

Some of the projects already circulating around the venue reflected how far builders are stretching the category. One team called Praxis was working on blockchain-connected drones controllable through smartphones, what the brothers described as “the next Palantir on the blockchain.” Another startup was building what they called “hyper-intelligent AI,” software designed to turn text prompts into physical 3D objects. “You could put in a prompt and say, ‘Build me a microscope,’ and it will actually build it for you,” Phil said. “It’s like the next phase of taking ChatGPT from something informational into something embodied.”

The winners:

The judges rewarded projects that pushed AI agents beyond chatbots and into real-world coordination, automation and commerce, whether through hardware, payments infrastructure or consumer-facing apps. Across the different sponsor tracks, the winning teams reflected the broader shift underway at this year’s hackathon: developers were no longer just building crypto tools, they were building products meant for everyday use. Prizes differed per track, and are still pending on how they will be divied up in each category.

Kickstart Track ($50,000):

First place: FlyPraxis

Taking the top spot in the Kickstart track was FlyPraxis, a real-time drone intelligence platform designed for military operators. The team pitched the project as “Palantir, but in real time,” using AI-powered coordination and live battlefield intelligence to manage autonomous drone systems.

Second place: HIIE

HIIE placed second with a platform that turns text prompts into fully buildable hardware products. Using AI agents to manage everything from physics calculations and component sourcing to 3D CAD generation and assembly documentation, the startup aimed to compress months of hardware prototyping into a single workflow.

Third place: Clan World

Clan World rounded out the top three in the Kickstart track, joining a broader wave of teams experimenting with AI-native coordination and community-driven applications.

Solana Mobile Track ($30,000 + $75,000 worth of Solana phones)

First place: Parabola

In the Solana Mobile track, first place went to Parabola, a decentralized prediction and estimation market built on Solana. The platform allows users to speculate on real-world events through a distribution-based AMM model designed for mobile-native trading experiences.

Second place: Snakr

Snakr took second place with an AI-powered food intelligence app that lets shoppers scan products to identify potential health risks, FDA recalls and ingredient concerns. Users can also contribute missing product information and earn Solana-based rewards in return.

Third place: Rhythym

Third place winner Rhythym focused on productivity and accessibility, building a mobile routine-support app aimed at helping users with executive dysfunction complete daily tasks. The app integrates with Solana’s Seeker phone, Nova 2 Lite and x402 infrastructure to create AI-assisted workflows.

Coinbase / AWS Track ($45,000)

First place: Dairy Price API x402

The Coinbase and AWS track centered heavily on AI-agent payments and autonomous commerce. The winning project, Dairy Price API x402, built a pay-per-call commodity pricing and forecasting service that allows AI agents to access dairy market data without traditional API keys. Payments are settled directly in USDC through x402 on Base.

Second place: AgentPay

AgentPay placed second with a payment coordination system that gives users one-tap approval over AI-agent transactions while using AWS-powered risk validation to ensure agents spend funds responsibly.

Third place: Giggy

Giggy took third place for building a marketplace where users can hire AI agents to perform research tasks. Payments are locked in crypto escrow on Base, while the agents themselves can pay for premium APIs through x402-powered transactions.

Runner up: Chainlens

Chainlens focused on trust and verification for autonomous systems, building an x402-compatible layer that connects AI agents to verified APIs and only releases payment once responses are authenticated.

Read more: AI-powered agents dominate the EasyA x Consensus Hong Kong hackathon