The Web3 + AI Daily: On NEAR's Token-Powered AI Economy
NEAR is turning into the quintessence of what decentralized AI has to offer.
Hello, everyone, and welcome to all newly joined subscribers!
This is the 79th edition of The Web3 + AI Daily - your definitive guide to the intersection of blockchain and AI. Today, we're once again talking about NEAR Protocol and NEAR AI - the blockchain that fully integrates AI as a first-class citizen, and powers confidential and verifiable AI at scale.
Thank you for being here! Let’s dive in.
NEAR’s Token Model for the AI Economy
In my previous newsletter, I told you about NEAR Protocol‘s launch tying the $NEAR token with NEAR AI‘s inference and agent offerings. Today, after Illia Polosukhin‘s appearance at the Bankless podcast, we have a clearer picture of what this means for NEAR in the long term. Here’s the gist of that conversation.
NEAR is building an entire technology stack to deliver sovereign AI, with NEAR AI built on top of NEAR Protocol’s blockchain and core cryptography primitives like intents. The goal is twofold: guaranteeing confidentiality:
“What you get as a developer is an end-to-end confidential AI inference. There’s nobody else who can access what queries or prompts you’re putting in.”
And ensuring verifiability - something the world beyond Web3 hasn’t cared much about, but it’s high time it started taking into consideration. Why? Because without the ability to verify, we have no way of knowing what model is actually running in the background and how it might affect our entire application / business / agentic system. In short, we’re entirely at the mercy of our model provider, just like every client of OpenAI, Anthropic, and Google is at the moment. In a single sentence Polosukhin illustrated this perfectly:
“If you want to manipulate a billion people right now into believing something, the easiest way is to get a job at OpenAI and modify the system prompt.”
The NEAR AI Stack
As David Hoffman commented, NEAR is like no other chain - it combines the properties of a general-purpose blockchain like Ethereum with those of an application-specific network, where AI is a first-class citizen. And that’s true. Over the years, NEAR has been building seemingly fragmented products, but this latest upgrade unifies its stack into a cohesive and fully-integrated system. It consists of:
IronClaw - NEAR’s agentic harness, born as the privacy-focused alternative of OpenClaw. It offers cryptographically secure agentic infrastructure where user data remains secured in encrypted vaults.
An agent marketplace, relying on the same infrastructure to guarantee that everything agents do can be verified and trusted.
NEAR AI Cloud - NEAR’s confidential and verifiable inference cloud, running inside TEEs (trusted execution environments), and serving various open-weights models. It currently relies on proprietary GPUs, whereas an underlying market involving third-party compute providers is being developed, too. NEAR AI Cloud has already attracted notable clients like Venice.ai, Brave, and The Government of Bermuda.
Intents - I’ll use Messari‘s definitions to explain what intent are:
NEAR uses intents to manage the above-mentioned compute market and route the compute capacity distributed all over the world. Since a compute buyer doesn’t always know exactly what type of GPUs they need, they only need to declare their rough estimates and expected workloads for solvers to then find the required resources and provide them to the client.
Intents also allow users to make compute into an asset, trade it, and create liquidity for the GPU hours itself, not just for inference.
$NEAR tokenomics - As Polosukhin said, the goal is to enable users to pay everything across the stack in $NEAR, “but we’re not there yet.” However, now that $NEAR stakers can continuously access AI capacity, NEAR already feels as one vertically integrated and unified infrastructure.
By staking $NEAR, holders can pay for both inference at NEAR AI Cloud and subscription to IronClaw’s agentic services. This is made possible by redistributing the yield generated from staking to compute providers. As a result, GPUs cluster owners are incentivized to join, as the more inference they perform for the cloud, the more $NEAR emissions will be staked to them.
NEAR blockchain - the blockchain is what operates underneath the entire stack and facilitates security for every component of it:
“Blockchain is the backend of this whole system, that ensures trust, identity, settlement, and coordination. It’s the core security, the root of trust, where the value settles.”
NEAR’s unique design can cater to very niche use cases that no other blockchain can power, including fully autonomous agentic businesses:
“You can call AI inference from inside a smart contract - a transaction can pause, wait for AI inference, and unpause and continue. There are very specific use cases where you’d want to use this, but the system is really integrated to allow for just that. The reason why it works is because you need verifiable inference that comes back into your blockspace. If you have money involved and a smart contract waiting to implement, you can’t rely on unverified inference. You have to verify end-to-end the supply chain of the prompt, the delivery, the compute, and the model.”
All Pieces Supporting Each Other
As you can see, NEAR is well-equipped to deliver a consolidated and coherent AI system, optimized for privacy. Most importantly, this latest tokenomics update makes it possible for the $NEAR token to actually capture value from AI:
“[The $NEAR token] needs to be able to provide the ability to access AI, which is the staking. And it needs to capture the transactions, the volume, the interactions. This is the intents. To be a true store of value, it needs the sovereignty, security, blockspace, and programmability provided by the blockchain.”
With the token becoming the utility for AI, the intents bootstrapping a global market for everything AI could do, and the blockchain securing the whole system, NEAR showcases what decentralized AI is all about.
Web3 + AI Readings & Conversations
Pacing The AI Frontier Is Simply Not Possible
Since we’re on the topic of NEAR and its co-founder, I thought I’d conclude today’s edition with a publication by Polosukhin which deeply resonated with me. The article was written as a response to a call by AI scientists and staffers for a “deliberate pacing” of AI frontier research.
Let me start with some context: at the end of July, 1,376 employees of frontier AI companies like OpenAI, Anthropic, and Meta signed an open letter to urge the US government to intentionally slow down AI development. You all need to read the exact request of the signatories, so I’ll quote it here:
“We request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.”
My immediate thoughts after seeing this were:
We have a saying in Bulgarian: “You can’t stop the forest from coming into leaf,” and it feels particularly apt here. How can the government “slow down” AI research? Is it going to ban it altogether, (and if yes, who and how will enforce that ban), or just prevent some actors from performing it?
It is particularly arrogant for US companies to ask the US government to impose rules internationally. Moreover, these are the same companies that created the most powerful AI models in existence, and the same US government that persistently supports and finance them. May I add - the same US government that treats AI as strategic national power. From that point of view, such a petition sounds more like an effort to stifle international competition than an expression of a genuine concern.
In his blog post, Polosukhin shares similar views, which motivated him to not sign the open letter.
“These frontier labs say they’re nervous that recursive self-improvement is moving us rapidly towards ASI, which we aren’t ready for, but they can’t slow down because the pressure of competition is too intense. This is an entirely wrong framing of the problem in front of us. We know major frontier labs have lobbied against open source to minimize competition and now they appear to be doubling down. I agree with the letter signers that there needs to be governance over AI, but I don’t agree that regulation by the government(s) is the answer. There are no convincing examples from history of government regulation effectively slowing down existentially risky competition between companies.”
Instead of imposing restrictions which won’t work anyway, Polosukhin urges AI scientists and researchers to cooperate and explore creative solutions to properly manage and govern AI. He also reaffirms the potential of existing decentralized solutions to ensure verifiability and to incentivize people to address AI safety concerns at a system level.
“We can do this today using existing decentralized infrastructure that can create a cooperative environment to work on verifiable inference. We can build an open source evaluation system that is verified by everyone but can score and red team models without accessing weights (vs. safety opaquely claimed by the author of the model). We can open up the ability to source community-driven alignment examples. We can create a value function for alignment that is aligned with the success of individuals rather than the companies’ profits, create economic insurance and slashing systems, and more.”
Read the full article and let me know your thoughts.
Thank you for reading! I hope you found this article insightful.
If you’d like to keep up with the latest developments at the burgeoning intersection of blockchain and AI, I invite you to subscribe.
And if you find value in my writing, consider supporting the newsletter. I’d greatly appreciate it.
The Web3 + AI Book Club is live on Fable. We’re currently reading Karen Hao’s ‘Empire of AI.’ Follow the link below and come join our literary community.
I'm looking forward to connecting with fellow Crypto x AI enthusiasts, so don't hesitate to share your thoughts here or reach out on LinkedIn, Instagram, or Firefly.social.
Disclaimer: The content of this newsletter is provided for informational and educational purposes only. Nothing contained herein should be construed as financial advice or as a recommendation to buy, sell, or hold any of the companies or assets mentioned.
Please don’t take my views at face value. Instead, do your own research (DYOR), think critically, and share your perspective so we can challenge ideas, learn from one another, and arrive at better conclusions through thoughtful discussion.