If you've ever fumbled through a seed phrase, misread a contract address, or watched a transaction confirmation bar crawl for three minutes, you've experienced what the crypto industry politely calls "the UX problem." The standard response has always been: we'll fix it. Better wallets. Simpler onboarding. Abstract the keys away. Make it feel like PayPal.
Nikil Viswanathan, co-founder and CEO of Alchemy — one of the most important blockchain infrastructure companies operating today — has a different read entirely. His argument isn't that crypto is hard to use and needs to be simplified. It's that crypto was never really designed for human users in the first place. The real end user, whether or not anyone planned it that way, is the AI agent.
That's not a dismissal of crypto. It's a reframe that has significant implications for where the next wave of genuine adoption actually comes from.
What Makes Crypto Awkward for Humans Makes It Perfect for Machines
Viswanathan's case is straightforward: the properties that make crypto frustrating for retail users are exactly the properties that make it ideal infrastructure for autonomous software agents.
Private keys are a constant headache for humans. Agents don't have that problem — they can generate, store, and use cryptographic keys without friction or anxiety. Reading contract bytecode or navigating a DEX interface without guidance is opaque to most people. An agent can parse that data directly, without a simplified front end.
More importantly, the structural properties of blockchain networks — borderless operation, 24/7 availability, no intermediaries, fully programmable settlement — line up almost exactly with how AI agents need to move value. An autonomous agent running a logistics optimization task at 2 a.m. on a Sunday, coordinating with counterparties across four countries, doesn't care that the New York Stock Exchange is closed. It doesn't need a correspondent bank relationship. It just needs rails that are always on and always accessible.
That's precisely what crypto provides. The settlement layer isn't the problem. The human interface was always the mismatch.
The Real Adoption Curve May Not Be Consumer-Facing
This matters because it reshapes how you should think about crypto's mainstream adoption arc.
The conventional narrative runs something like this: crypto succeeds when your parents can use it without reading a tutorial. When you can send Bitcoin the same way you send a Venmo payment. The benchmark is always human usability.
But if Viswanathan's framing is right, that benchmark may be measuring the wrong thing. The more meaningful signal would be how many AI agents are settling payments, managing treasury balances, or executing conditional micropayments on-chain right now — and how quickly that number is growing.
We don't have clean public data on agent-driven transaction volume, but the infrastructure investment tells you something. Companies like Alchemy are building developer tooling that makes it easier to deploy AI agents on top of blockchain networks. That's not consumer fintech. That's machine infrastructure.
The implication for retail crypto investors is worth sitting with: the killer app for crypto may not look like an app at all. It may look like a background process.
What This Means for the Infrastructure Layer
If AI agents become a primary transaction class on blockchain networks, the infrastructure demands shift considerably.
Speed and finality matter more than they already do. An agent executing a high-frequency task can't wait fifteen seconds for a block confirmation. This is already driving demand for Layer 2 networks and high-throughput chains, and it connects directly to the Ethereum Foundation's ongoing push to have L1 and L2 function as a unified system rather than competing alternatives. The stated goal — scaling Ethereum as a "cohesive system" — looks more urgent if the anticipated user base is measured in millions of software agents rather than millions of human wallets.
Cost predictability also becomes critical. Humans can absorb the occasional $40 gas fee with frustration. An agent running thousands of micro-transactions per hour cannot. Fee markets, gas abstraction, and stable settlement costs are no longer just quality-of-life improvements. They're table stakes for machine adoption.
And identity starts to look different. Human identity in crypto is already a mess — pseudonymity, wallet addresses that mean nothing to outsiders, KYC bolted on awkwardly at the exchange layer. Agent identity is a different and arguably harder problem. How do you verify that a contracting agent is authorized to commit funds on behalf of an institution? How do you revoke credentials from an agent that's been compromised? These are infrastructure questions that don't have clean answers yet, but they're coming.
The Stablecoin Layer Underneath It All
There's another piece of this that doesn't get enough attention: agents need a stable unit of account. An agent managing treasury operations or cross-border payments can't effectively do its job if its settlement asset swings 8% in a day.
This is part of why stablecoin infrastructure has become so critical, and why the volume numbers are striking. Stablecoin transaction volume reached $33 trillion in 2025, exceeding global credit card volume, according to Ripple. That's not a retail adoption story. That's a story about institutional and automated flows moving through rails that work around the clock without correspondent banking overhead.
The trend toward multi-stablecoin operations — institutions running USDT, USDC, EURC, and local-currency stablecoins simultaneously, choosing assets based on corridor and counterparty — is exactly the kind of programmable, context-sensitive behavior that agents can manage well and humans cannot do efficiently at scale.
Where This Leaves Retail Investors
If the next wave of meaningful crypto adoption is agent-driven, a few things follow.
Network activity metrics deserve more scrutiny. Raw transaction counts will become less useful as a signal of genuine adoption if a large share of on-chain activity is automated. The composition of transactions — size, frequency, contract type — may matter more than volume alone.
Infrastructure tokens gain a clearer thesis. Networks that can credibly serve as settlement rails for high-frequency, low-latency agent activity have a more concrete demand story than those whose value proposition is primarily speculative. This isn't a recommendation, but it's a more grounded frame than "number of wallet addresses."
And the consumer UX problem doesn't disappear — it just gets deprioritized. Retail users still need better interfaces. But if the growth engine over the next several years is agent-driven activity rather than consumer adoption, expecting human usability improvements to drive price action may be waiting for the wrong catalyst.
The Honest Takeaway
Viswanathan's argument isn't a prediction that humans will stop using crypto. It's a reframe of where the design center of gravity actually sits. Crypto infrastructure — its cryptographic primitives, its 24/7 operation, its programmable settlement — was built in a way that happens to align well with how autonomous software agents need to move value at scale.
That's not a marketing pitch. It's an architectural observation. Whether the market has priced it correctly is a separate question, and a harder one. But understanding what the technology is actually optimized for — rather than what we wished it was for — is a better starting point than most of the narratives currently on offer.