At DePIN Day Singapore, Lane Rettig, Head of Research at the NEAR Foundation, offered a clear-eyed reflection on the state of decentralized AI — and where it’s heading next.
A former Ethereum core developer and co-founder of SpaceMesh, Rettig has spent nearly a decade thinking about how technology can build better human institutions. “It was never the financial side of crypto that excited me,” he said. “It’s always been the social side — the chance to create systems that people can actually trust.”
Today, that mission sits at the intersection of AI, blockchain, and DePIN — three forces shaping the next era of digital infrastructure.
Scalable blockchains to AI infrastructure
Few people know that NEAR itself began not as a blockchain project but as an AI experiment. In 2017, founders Illia Polosukhin and Alex Skidanov set out to build coding tools powered by machine learning — long before “AI copilots” became a mainstream concept. When scalability proved a bottleneck, they pivoted to building a new blockchain to support that vision.
Six years later, NEAR is circling back to its roots. The foundation now aims to create a decentralized AI cloud, uniting compute, governance, and cryptography into one interoperable layer.
Rettig broke down what that looks like in practice:
- Sharded architecture for near-infinite throughput — “one million transactions per second is within reach.”
- Chain abstraction, enabling a single NEAR account to interact across multiple blockchains.
- Intents and solvers, allowing autonomous AI agents to act on behalf of users — from transferring assets to booking flights.
- House of Stake, a decentralized governance DAO.
- And finally, NEAR AI Cloud, a verifiable, privacy-preserving compute layer that allows users to run AI models securely on decentralized infrastructure.
Together, these elements form NEAR’s answer to a growing question in Web3: how do you make AI not just powerful, but trustworthy?
Why decentralized AI can’t out-compete Big Tech and shouldn’t try
Rettig’s talk took a critical turn when he addressed what he called “the industry’s biggest delusion” — the idea that decentralized AI can rival the hyperscalers on cost or efficiency.
“I don’t think crypto will ever beat centralized systems in raw performance,” he said. “Those $10-billion data centers have GPU interconnects running at 10 terabits per second — ten thousand times faster than my home internet.”
Instead, he argued, Web3’s strength lies elsewhere: in privacy, transparency, and sovereignty — the very attributes mainstream AI has sacrificed.
Today’s dominant AI models serve a narrow slice of the world: affluent, tech-savvy users and corporations who value speed and convenience above all. But, Rettig pointed out, “there’s a massive population that’s not being served — and in some cases, actively excluded.”
That’s where decentralized AI can thrive.
Three personas that define the decentralized AI market
Borrowing from classic market segmentation, Rettig introduced three archetypes that capture the emerging landscape:
- Priya — The D-Sci Builder A young researcher who believes in open science and collective progress. She’s not a cypherpunk but an openness maximalist — someone who wants AI models to be open-source, community-governed, and accessible to all. For her, NEAR’s focus on verifiable model training and community governance could unlock a new paradigm: models where contributors can cryptographically prove their data was used and receive compensation when it’s applied downstream. “Imagine scientists proving their work was part of a model — and getting paid for it,” Rettig said. “That’s the kind of future we’re building toward.”
- Alex — The Consultant A professional obsessed with efficiency and cost reduction. He uses ChatGPT and corporate AI tools because they’re fast and reliable. “We’re not going to win Alex,” Rettig said. Competing here would mean chasing the same incentives as Big Tech — and losing what makes Web3 unique.
- Satoshi — The Cypherpunk A privacy maximalist who values control over convenience. He runs Bitcoin nodes, uses Signal, audits code, and wants verifiable proofs that no one can access his data. “This,” Rettig emphasized, “is our user.” For Satoshi — and for industries like finance and healthcare bound by strict compliance — NEAR’s confidential, verifiable compute provides exactly that: mathematical proof that nothing leaves the secure enclave. Through the NEAR AI Cloud, users can run AI tasks and receive a cryptographic proof for every interaction — verifiable by hardware vendors like NVIDIA or Intel. “Even if you don’t control the hardware,” Rettig explained, “you can still have privacy guarantees.”
Redefining “trust” in AI
If traditional AI optimizes for performance, Rettig envisions decentralized AI optimizing for trust — where confidentiality, provenance, and community oversight are built into the architecture itself.
“Let’s stop pretending our superpower is efficiency,” he said in closing.
“Our superpower is verifiability. It’s privacy. It’s governance. It’s giving people — and societies — control over the systems that increasingly shape their lives.”
As NEAR Foundation brings its AI vision full circle, it’s not just reviving the project’s origins — it’s reframing the narrative around what kind of intelligence the decentralized web should aspire to build.
