AI agents will be top blockchain users: NEAR co-founder

Crypto keeps begging for "mass adoption," and the universe keeps replying: Sure, but not for humans.

That was the subtext of NEAR Protocol$1.4193 co-founder Illia Polosukhin's message this week: the next big wave of blockchain users will be AI agents, not people hunched over mobile wallets, squinting at transaction hashes like they are reading ancient runes.^[1] At roughly the same moment, the market was doing its usual macro shuffle, with Bitcoin$62,424.01 around $68,216, Ethereum$1,686.33 near $1,976, Solana$79.10 about $84.91, and XRP$1.1041 around $1.36. Prices move, narratives rotate, and the industry still can't decide whether it is building a financial system or a complicated captcha.

Polosukhin's argument cuts through that ambiguity with a simple claim: AI becomes the interface layer for the internet, and crypto becomes one of the rails that AI uses, quietly, constantly, and mostly out of sight.^[2]

OKX unveils OnchainOS, a toolkit that lets AI agents trade and swap autonomously on-chain

Enjoy articles without ads?

The claim, plainly: AI becomes the interface, crypto becomes the plumbing

Polosukhin's core prediction is not that humans will fall in love with block explorers. It is that they will stop needing to see them at all.

If AI agents become the default way people "use" the internet, then it follows that agents will also become the default way people "use" blockchains. Users will express intent in plain language (or through an app that does it for them), and software will do the rest: selecting a chain, sourcing liquidity, signing transactions, tracking receipts, and reconciling balances.

That matters because crypto UX is still, bluntly, not normal-person software. Wallet setup, seed phrases, gas fees, network switching, transaction hashes, and irreversible mistakes are not "early adopter friction." They are structural barriers.

Polosukhin's bet is that AI smooths that surface by turning crypto into a backend. The human asks for an outcome, the agent executes the steps.

Why this lands now: AI is already eating "interface" work

AI's rapid spread has been obvious in daily workflows: drafting text, writing code, summarizing meetings, planning travel, and handling customer support. The shift is not just "AI is smart," it is "AI sits between you and the messy parts."

That interface role is exactly where crypto has struggled. Even many crypto-native users treat wallets like fragile containers that might explode if you click the wrong button. Asking the next billion users to do key management is like asking the next billion drivers to build their own transmission, because it's "more decentralized."^[3]

So the timing checks out. AI agents are getting embedded into products people already use. Crypto is still hunting for the next breakout moment. Polosukhin is essentially saying: the breakout might come, but it will look boring from the outside, because the users will be software.

The practical upside: intent-based transactions and invisible complexity

Polosukhin's view implies a shift from user-driven clicks to intent-driven execution.

"Intent" here means the user describes what they want, not how to do it. Example: "Move $200 to savings that earns yield, but keep it liquid," or "Buy the cheapest flight and hotel, pay from my stablecoin balance, and send receipts to accounting." The agent then decides whether to swap assets, bridge across chains, or choose a payment route. The blockchain becomes a settlement layer that the user does not need to micromanage.

This is also where Polosukhin's "wallets and transaction hashes fade" idea lands. Wallets do not disappear, but they stop being the primary interface. Transaction hashes still exist, but they become machine-level receipts, not user-level artifacts.

That is a real UX improvement, and it could finally align crypto with how normal consumer products work: people ask for outcomes, systems handle the implementation.

Takeaways (so far)

AI agents can hide crypto's worst UX, which is a more plausible adoption lever than hoping everyone learns key management.
"Primary users" may shift from humans to software, with humans acting as policy setters and approvers.
Blockchains become rails, not destinations. That is good for usage, and bad for tribal marketing.

The commerce angle: agents do not just transact, they negotiate

Once software is transacting, it does not stop at "click to buy." It compares options, splits routes, arbitrages fees, and negotiates terms.

That aligns with a broader theme bubbling across fintech: agent-driven commerce, where software can search, select, and pay on behalf of the user or a business. Some industry leaders outside crypto have floated similar forecasts: a surge in machine-to-machine payments, subscriptions, and procurement, with humans overseeing constraints rather than approving every step.^[4]

Crypto fits here because blockchains are, at their best, programmable settlement networks. Stablecoins are the obvious payment unit, and smart contracts can encode conditional actions. If agents become high-frequency economic actors, crypto's always-on rails start to look less like a hobby and more like infrastructure.

The irony is hard to miss: crypto's long-awaited "real users" might not have feelings about decentralization at all. They will just care about latency, reliability, and cost.

Paradigm Raises $1.5B Fund to Expand Beyond Crypto Into AI and Robotics: WSJ

The hard part: agents that can spend money are a security problem

Letting an AI agent "handle" crypto is also a perfect way to speedrun every risk the industry has spent a decade collecting.

Key questions Polosukhin's thesis forces into the open:

1) Who controls the keys, and how is authority scoped?

An agent that can sign transactions is effectively a hot wallet with a personality. Without strict permissioning, it becomes a single point of failure with more ways to be tricked. The workable model is likely scoped authority: limits by amount, asset, contract address, time window, and allowed actions.

2) What happens when an agent is manipulated?

Prompt injection is not theoretical. If an agent reads messages, webpages, invoices, or "support tickets," attackers will try to feed it instructions that look legitimate. On-chain finality means "oops" is not reversible.

3) Can compliance and user protection keep up?

If agents transact across protocols and jurisdictions, regulators will ask who is responsible when something goes wrong: the user, the agent developer, the wallet provider, or the protocol. The industry does not have clean answers, and "code is law" is not a compliance strategy.

4) Will blockchains handle agent-scale transaction patterns?

Agent usage could mean more frequent, smaller actions. That stresses throughput and fee markets. It also makes "which chain" a practical decision again, not a branding contest.

What this means for NEAR, and for everyone else

Polosukhin's position is also a strategic pitch: if AI is the new interface, then blockchains that make "agent UX" easier have a shot at capturing real activity.

That means prioritizing:

Account abstraction and better key management (so agents can operate safely under constraints)
Readable permissions and policy layers (so humans can audit what the agent is allowed to do)
Interoperability (because agents will route around chain tribalism)
Privacy and confidentiality options (because not all commerce should be globally legible by default)

NEAR Protocol$1.4193 has marketed itself around usability and developer experience, and it has signaled interest in infrastructure that supports more complex applications. The agent thesis is consistent with that direction, even if the industry is still early in proving it at scale.

What to watch next

The "AI agents as primary users" prediction will not be validated by a single app launch or a conference demo. Look for measurable signals:

Wallets shipping agent permissions that are actually granular, not just "approve all."
Stablecoin payment flows that run end-to-end through agents, including invoicing, reconciliation, and refunds.
On-chain activity that shifts toward many small, automated transactions, rather than human-driven bursts around hype cycles.
Security incidents specifically tied to agent workflows, including prompt manipulation and permission abuse (because of course attackers will go there first).
Regulatory guidance on agent-mediated transactions, especially around custody, liability, and consumer protection.

Polosukhin might be right that the next wave of blockchain users will not be humans. The more uncomfortable truth is that crypto will have to grow up operationally if it wants software to trust it. Agents will not tolerate nonsense. Humans sometimes do.

NEAR Co-Founder Predicts AI Agents—not Humans—Will Become Blockchain’s Main Users

NEAR rallies 17% after launching ‘Confidential Intents’ privacy layer to curb MEV and front-running