NEAR Protocol's 'Invisible Crypto' Gambit: How a Transformer Co-Author Is Betting That AI Agents — Not Humans — Will Drive the Next Billion Blockchain Transactions
The co-author of "Attention Is All You Need" — the paper that spawned ChatGPT, Gemini, and the entire large language model revolution — believes that the future of crypto is not about getting more humans to use wallets. It is about making crypto so invisible that neither you nor the AI agent booking your flights, managing your portfolio, and paying your bills ever needs to think about it.
On February 23, 2026, NEAR Protocol launched near.com, a super app that bundles wallet management, confidential transactions, AI-powered insights, and cross-chain asset management into a single consumer interface. Within days, NEAR's token surged 40% in a week, Confidential Intents went live enabling private cross-chain swaps, and the NEAR Intents framework crossed $10 billion in all-time volume. This is not a typical protocol upgrade announcement. It is a full-stack thesis about what crypto becomes when AI agents outnumber human users on-chain.
From Sharded L1 to AI Blockchain: NEAR's Identity Pivot
NEAR Protocol started life in 2020 as a sharded Layer 1 blockchain optimized for developer experience. Nightshade sharding, human-readable account names, and a WebAssembly runtime attracted a respectable developer community, but NEAR struggled to differentiate itself from the Solana-Ethereum-Avalanche triad in terms of mindshare and TVL.
The pivot began when co-founder Illia Polosukhin — whose Transformer architecture research at Google literally created the foundation for modern AI — started articulating a thesis that most crypto observers initially dismissed: AI agents, not humans, will be the primary users of blockchain.
The logic runs as follows. As AI systems evolve from chatbots into autonomous agents that book travel, manage investments, hire freelancers, and negotiate contracts, they need three capabilities that traditional payment rails cannot provide:
- Programmable payments — agents need to send and receive micropayments autonomously, 24/7, without human approval for each transaction
- Cross-chain interoperability — agents interacting with services across different blockchains need seamless asset movement without manual bridging
- Privacy — agents managing financial strategies cannot broadcast their intentions to a public mempool where MEV bots front-run every transaction
NEAR's 2026 product roadmap addresses all three. And the near.com super app is the consumer-facing expression of that thesis.
Inside near.com: Making Blockchain Disappear
The near.com app launched with a deceptively simple promise: use crypto without knowing you are using crypto. Users manage Bitcoin, stablecoins, NFTs, and tokens across multiple chains from a single interface. There are no gas fee prompts, no seed phrase ceremonies, no chain-switching dropdowns.
Behind that simplicity sits NEAR's chain abstraction stack, which has processed over $10 billion in cross-chain swap volume through the Intents framework. In the last 30 days alone, NEAR Intents handled $2.15 billion across 541,075 unique addresses — roughly 20% of its all-time volume concentrated in a single month.
The app layers AI on top of this infrastructure. AI agents embedded in near.com analyze user activity, surface insights, identify portfolio risks, and recommend financial tools. Polosukhin's vision extends further: these agents will eventually handle travel bookings, email management, and online purchases — tasks that require the ability to send and receive payments autonomously.
This is the "invisible crypto" thesis in practice. The blockchain becomes plumbing. The AI agent becomes the interface. The user never sees a transaction hash.
Confidential Intents: Privacy as Infrastructure, Not a Feature
On February 25, 2026 — two days after the super app launch — NEAR rolled out Confidential Intents, and the market responded immediately. The NEAR token jumped 17% in a single day, extending a roughly 40% weekly rally that outpaced the entire privacy token sector.
Confidential Intents allow users to execute cross-chain transactions — swaps, deposits, transfers — within a restricted-visibility execution environment powered by NEAR's private shards and Trusted Execution Environments (TEEs). Users toggle between a standard "Main Account" and a "Confidential Account" directly in the near.com app.
The practical implications are significant. In standard DeFi, every pending transaction is visible in the public mempool. MEV bots scan these transactions and front-run profitable trades, extracting value from ordinary users. Confidential Intents hide transaction details during the settlement phase, making front-running, sandwich attacks, and strategy copying structurally impossible within the private execution environment.
But the deeper strategic play is not about protecting human traders — it is about enabling AI agents. Autonomous agents managing portfolios, executing arbitrage strategies, or settling cross-border payments cannot operate effectively if their every move is visible to adversarial bots. Confidential execution is not a nice-to-have for the agentic economy. It is a prerequisite.
The NEAR AI Stack: From Cloud to Agent Marketplace
NEAR's AI ambitions extend well beyond wallet UX. The protocol is building a full-stack AI infrastructure:
NEAR AI Cloud launched version 26.1 in February 2026, hosting open-source models like GLM-4.7 — a 355-billion-parameter Mixture-of-Experts model with 32 billion activated parameters — inside GPU Trusted Execution Environments. This means AI inference runs with end-to-end encryption, protecting both the model and the user's data throughout the computation lifecycle. HOT Pay enables crypto-native payments for AI inference, allowing developers and agents to pay per computation in NEAR tokens.
The NEAR AI Agent Market, launched on February 5, 2026, creates an on-chain marketplace where users post tasks with specified budgets, AI agents bid competitively, and the selected agent executes the work autonomously before receiving payment in NEAR upon completion. This is essentially a decentralized labor market for AI, with blockchain serving as the trust and settlement layer.
HOT Wallet, built by HOT Protocol, serves as the primary consumer wallet for the NEAR ecosystem, supporting multi-chain transactions through MPC (Multi-Party Computation) wallets. HOT Wallet nodes are hosted by trusted validators including Everstake, Aurora, and HAPI, providing the security infrastructure that makes the super app's seamless UX possible.
The Competitive Landscape: Who Else Wants to Be the AI-Crypto Super App?
NEAR is not alone in pursuing the AI-blockchain convergence thesis, but its positioning is uniquely credible.
Internet Computer (ICP) is pursuing a similar "world computer" narrative with its Mission 70 deflationary pivot and a sovereign AI cloud partnership with Pakistan. But ICP's approach focuses on hosting full applications on-chain rather than building consumer-facing super apps, and it trails NEAR in developer mindshare and cross-chain interoperability.
Bittensor has carved out a niche in decentralized model training and inference incentives, but operates primarily as infrastructure for AI researchers rather than a consumer product. Its subnet model creates powerful incentive alignment for compute providers but does not address the "last mile" UX problem that near.com tackles.
LINE NEXT's Unifi and X Money have both attempted the crypto super app model from different angles — messaging-first and social-media-first respectively — but neither has shipped the combination of chain abstraction, confidential transactions, and AI agent infrastructure that NEAR has assembled.
Polosukhin's background as a Transformer co-author gives NEAR a narrative advantage that is difficult to replicate. When the person who helped invent the architecture behind ChatGPT says AI agents will be blockchain's primary users, the market listens. The 40% weekly price surge after the near.com launch suggests that this narrative premium is real and tradeable.
The $10 Billion Question: Can Intent-Based Architecture Scale?
NEAR's Intents framework represents a fundamental rethinking of how users interact with blockchains. Instead of manually constructing transactions, specifying gas parameters, and selecting bridges, users express an intent — "I want to swap 100 USDC on Ethereum for SOL on Solana" — and the system finds the optimal execution path.
The numbers suggest this model is gaining traction. NEAR Intents has processed $10 billion in all-time swap volume across 120+ assets, with monthly volume now running at approximately $2.5 billion. The Intents Widget allows third-party applications to embed this cross-chain functionality directly, turning what would normally require multiple bridge transactions into a single user interaction.
For 2026, NEAR's Infrastructure Committee plans to scale chain abstraction through an expanded MPC network and multi-chain verifiable execution via TEEs. The integration with Starknet expands NEAR Intents' cross-chain reach into the zero-knowledge rollup ecosystem, and Cardano integration brings another major L1 into the fold.
The question is whether intent-based architectures can maintain performance and security as they scale to handle the transaction volumes that a true "agentic economy" would generate. If millions of AI agents are settling micropayments, executing trades, and managing portfolios simultaneously, the infrastructure requirements are orders of magnitude beyond what any blockchain processes today.
What This Means for the Industry
NEAR's bet on invisible crypto reflects a broader realization sweeping the industry: the winner of the next cycle may not be the chain with the most TVL or the fastest TPS, but the one that disappears most completely into the background of users' digital lives.
Three implications stand out:
The UX endgame is abstraction, not education. For years, the crypto industry has tried to educate users about wallets, gas fees, and private keys. NEAR's approach concedes that this is a losing strategy and instead hides the complexity behind AI and chain abstraction. If this works, it validates the thesis that crypto adoption is a design problem, not an education problem.
Privacy and AI are converging. Confidential Intents are not just about protecting human privacy — they are about enabling AI agents to operate in adversarial environments. As more economic activity moves to autonomous agents, the demand for confidential execution will grow exponentially. Chains that treat privacy as an afterthought will be structurally disadvantaged in the agentic economy.
The Transformer connection matters. Polosukhin's credibility as a co-author of the foundational AI paper gives NEAR a unique bridge between the AI and crypto communities. In a market where narrative drives capital flows, having a founder who literally helped create modern AI is a moat that no amount of marketing spend can replicate.
Whether NEAR's invisible crypto thesis ultimately succeeds depends on execution — shipping reliable AI agents, maintaining Confidential Intents security under adversarial conditions, and scaling the Intents framework to handle exponentially growing volume. But the direction of the bet is clear: crypto's next billion users might not be users at all. They might be agents.
Building for the agentic economy requires infrastructure that can handle cross-chain complexity at scale. BlockEden.xyz provides enterprise-grade RPC and API services across leading chains including Sui, Aptos, and Ethereum — the kind of reliable backend that both human developers and AI agents need to operate seamlessly across the multi-chain landscape. Explore our API marketplace to see what's available.