Does an Agent really need crypto payments?
Author: Lao Bai
I just spent half an hour reading this insightful piece by @Leozayaat.
It’s quite cognitively demanding, but offers fascinating insights, particularly regarding judgments about the future Agent Economy era.
Previously, it was widely believed that for economic activity in the Agent world, the best entry point for crypto was payments. In fact, a mainstream view within the industry over the past couple of years has been that blockchains are not built for humans, but for Agents, since owning a wallet is trivially easy and micropayments are highly suitable.
Just two days ago, DragonFly’s @hosseeb posted a Tweet, echoing similar points he made at 2049. He now acknowledges his previous assessment was flawed: Agents don’t need crypto; it’s us humans who do.
The "don't need" he refers to specifically targets payments.
Over the past year, companies like Visa and Stripe have been actively positioning themselves in Agent payments. The vast majority of Agent transactions go through traditional card networks and the legacy financial system, with crypto protocols like X402 accounting for a minuscule fraction.
Agents certainly require payment capabilities, but that doesn’t imply they need a brand-new payment network.
What if the new scenario for the Agent Economy isn’t payments, but something entirely different—say, financial markets?
Leo’s article focuses precisely on this premise.
Let me first briefly outline the original framework.
The Agent Economy may evolve through three stages.
Stage 1: Agents buying from humans. Agents purchase APIs, data, compute power, and other services on behalf of humans. The core challenge lies in enabling autonomous payments by Agents.
Stage 2: Agents buying from agents. Agents begin purchasing services from one another, shifting the problem from payment to fulfillment: how to guarantee delivery? The article proposes using Bonds / Collateral to enforce constraints on Agents. For Agents capable of infinitely generating identities, reputations are easily manipulated, whereas truly confiscatable collateral proves to be a far more reliable mechanism.
Stage 3: Disposable Markets. This is the most imaginative segment.
For example: An Agent temporarily requires 1,000 hours of GPU compute to complete a task but fears supplier delay. It can on-demand launch a market where other participants price the specific risk of "whether the supplier will deliver on time," providing capital to underwrite that risk.
The market does not need to be listed in advance or have market makers recruited beforehand. Provided there is genuine demand, it can simply seek out capital willing to assume the risk.
This implies that financial markets themselves could function as an on-demand infrastructure.
Here are my reflections after reading it:
1. Payments in Stage 1 currently carry a certain "falsification" vibe for X402, which is likely why Haseeb proactively "admitted his mistake." That said, it's still early days in the space, so declaring it unviable would be premature.
2. Reputation in Stage 2 is what ERC8004 is building. Like X402, current data suggests it is also leaning toward being "falsified." Leo’s original text contains detailed arguments for this perspective, which I won’t reproduce here. In short: "On-chain identity does not equate to trusted reputation; reputation records do not equate to verified transaction history; the cost for Agents to farm reputations is exceptionally low."
3. Stage 3 Disposable Markets—where Agents can on-demand create single-use markets to price specific risks and locate capital willing to absorb them.
This represents a completely new economic paradigm. Logically, it makes perfect sense in a silicon-based lifeform ecosystem, as their information retrieval, discovery, decision-making, communication, and pricing mechanisms differ fundamentally from human society, operating at efficiencies N orders of magnitude higher.
Whether it actually unfolds this way remains uncertain; it is merely a theoretical extrapolation. Yet it is at least logically consistent and opens up fresh space for imagination and possibility.
Especially given the recent launches of Grok Bot and Meta Muse, the evolution of the Agent Economy feels poised to accelerate. By then, novel forms of financial markets will undoubtedly emerge, extending far beyond mere payments.
If you have the time, I highly recommend reading the original article. It documents numerous experiments in economics and finance conducted by international labs using Agents, which are genuinely fascinating.