NEAR is a blockchain where using crypto feels like using a regular app. Your account is a readable name like alice.near instead of a 44-character string. You can sign up with an email address. One account can hold and move assets across multiple blockchains, and NEAR handles the coordination behind the scenes (which chain, which token, which fee) so you never see it.

The same design serves a second customer: AI agents. Because NEAR accounts are programmable and work across chains, an AI agent can hold an account, hold funds, and execute transactions on its own. It can pay for a service, swap an asset, or settle with another agent. NEAR is building for a future where your software does your errands, and the errands involve money.

The pitch in one line: people get a web2 experience, agents get open rails, and the blockchain machinery stays invisible to both.

Crypto asks a lot of a new user. Write down twelve secret words and never lose them. Copy a 44-character address perfectly, because one wrong character sends your money to nobody. Keep a different wallet for each chain, and hold each chain's special token just to pay its fees. I've done all of this, and I still triple-check every transfer with my heart rate up.

Why is it built this way? Because blockchains were engineered for security and decentralization first. Every safety guarantee pushed a piece of complexity onto the user: no central operator means no customer service, no undo button means no mistakes allowed, keys only you control means keys only you can lose. Then every chain developed its own conventions, so the coordination work between chains landed on the user too. The interface is one of several hurdles between crypto and mainstream use (regulation and volatility are others), but it is the one a builder can actually engineer away.

That is the specific hurdle NEAR was built to attack.

In 2017, eight researchers at Google published a paper called "Attention Is All You Need." It introduced the Transformer, the architecture underneath ChatGPT, Claude, and Gemini. One of those eight authors was Illia Polosukhin.

The next year, Polosukhin and systems engineer Alex Skidanov started NEAR as an AI company. NEAR.ai was teaching machines to write code, and the team needed a way to pay contributors around the world in small amounts. The payment rails they needed didn't exist, so they built them. The blockchain began as infrastructure for an AI project, which is why "NEAR is a blockchain for AI" is less a marketing pivot than a return to the founding premise.

The record since: mainnet running continuously since April 2020, more than 750 million transactions processed, roughly 100 million accounts created, and three of the top ten applications in Web3 by user count.

The NEAR token has three jobs. It pays transaction fees. It secures the network: people called validators lock up their own tokens as collateral and earn rewards for processing transactions honestly. And it votes on decisions about how the network runs.

Here is the simplest way to understand the economics. Every time someone uses NEAR, the fee they pay is destroyed. Gone from circulation, permanently. Meanwhile, the network creates a small amount of new tokens each year (about 2.5%) to pay the validators who keep it running. That rate used to be 5%; the community voted to cut it in half in late 2025.

So two forces pull on the supply. Usage destroys tokens, and validator pay creates them. If lots of people use the network, more tokens are destroyed than created and the token gets scarcer. If few people use it, the opposite happens. Owning NEAR is a bet that usage will grow. That is the whole value story, and it is more honest than most.

There is a second idea in NEAR's economic design worth knowing about, because it is quietly radical: a built-in payout to builders. As designed, 30% of every transaction fee goes to the developer whose app was used. Think about what that inverts. In web2, the platform takes the cut from the builder: Apple keeps 30% of App Store revenue, YouTube keeps roughly 45% of ad money. NEAR's design pays the 30% to the builder, automatically, with no platform in between. The industry calls this pattern protocol fee sharing, and it is a first attempt at a more equitable compensation model than the web2 platforms offer.

Here is the honest part: that mechanism is in NEAR's code but currently switched off. The protocol's own specification shows the payout set to zero, so today the full fee is destroyed. The equitable-compensation model is a design NEAR pioneered and hasn't yet turned on. A handful of smaller chains run live versions of the same idea, and that is a story I'll come back to in a future issue.

The strongest objection to NEAR goes like this: a friendly interface is not a defensible advantage, because competitors can copy it. Ethereum is already building toward readable, email-friendly accounts of its own. If the pitch is "crypto that feels like the web," every well-funded chain will eventually make the same pitch.

The real moat candidate is narrower than the interface. NEAR's cross-chain system, NEAR Intents, works like this: you (or an AI agent) state the outcome you want, such as "swap this asset for that one," and a network of competing market-makers, called solvers, races to execute it across chains at the best price. That is a two-sided network. More solvers means better prices, better prices attract more volume, and more volume attracts more solvers. A competitor can copy the code. Copying $22 billion of flowing volume and the solver network serving it is a different problem.

The precise answer to the moat question: the technology is copyable, the operating network is the moat candidate, and it only counts if adoption keeps compounding. Which brings us to the numbers.

When I evaluated NEAR last December, my framework asked for evidence of adoption and staying power, not just the ability to scale. Most of this list is newer than that evaluation. The evidence I asked for has started arriving.

My Gate test scores eight questions green, yellow, or red. Five greens and no reds to pass. Last December, NEAR scored two greens and six yellows: watchlist. Re-scoring against what's above:

Five greens, three yellows, no reds. NEAR passes the Gate.

Six months ago it didn't. What changed wasn't my standards; it was the evidence. The adoption December asked for showed up, the network kept running, and the community made a hard economic decision together. That is what a project earning its way through a framework looks like.

What I'm watching from here: whether Intents volume holds its pace through a full market cycle, and whether a second Brave-scale company puts NEAR's AI cloud in production. Those two signals carry more information than any price chart.

Whether NEAR belongs in your portfolio is your decision. What the Gate can tell you is that the gap between this project's story and its evidence, the gap that kept it on my watchlist in December, has closed in the direction of the evidence.

This is structural analysis. Nothing here is financial advice.

NEAR is a blockchain that gives people a web2-like experience: readable account names, email signup, one account across multiple chains. It handles the crypto coordination behind the scenes, and the same rails let AI agents hold accounts and transact directly. Mainnet has run continuously since April 2020.

NEAR Intents is NEAR's cross-chain transaction system. A user or AI agent states the outcome they want, like swapping one asset for another, and competing market-makers called solvers execute it across chains at the best price. Cumulative volume grew from about $3 billion in October 2025 to over $22 billion by mid-2026.

Illia Polosukhin, one of the eight co-authors of "Attention Is All You Need" (the 2017 paper that introduced the Transformer architecture behind ChatGPT, Claude, and Gemini), co-founded NEAR with systems engineer Alex Skidanov in 2018. It began as an AI company that built a blockchain because its work needed payment rails.

The token pays transaction fees, secures the network through staking, and votes on network decisions. Fees are destroyed when spent, permanently reducing supply, while about 2.5% in new tokens is created each year to pay validators. Heavy usage makes the token scarcer. A designed 30% fee-share to developers exists in the code but is currently switched off.

Next up: io.net, the AI supercomputer assembled from spare parts.