PANews columnist Lao Bai has revisited a framework attributed to @Leozayaat and shifted the discussion away from a narrower question — whether agents need blockchain payments — toward a broader one: if payments are not the main entry point, could the Agent Economy end up creating new financial markets instead.
The article says a common view in crypto has been that blockchains may fit agents better than humans because wallets are easy to create and machine-driven micropayments look like a natural use case. Lao Bai notes, though, that DragonFly's @hosseeb recently said on X that his earlier view was wrong and that agents do not need crypto, while humans do. In this context, the point was mainly about payments.
Lao Bai also writes that over the past year, companies including Visa and Stripe have been actively building around agent payments, while most of those payments still run through traditional card networks and the existing financial system. By comparison, crypto protocols such as X402 account for only a tiny share. His conclusion is restrained: agents do need payments, but that does not mean they need an entirely new payment network.
Three stages in the Agent Economy thesis
Under the framework summarized in the column, the Agent Economy could unfold in three stages.
Stage one: agents buying from humans
At the first stage, agents purchase services such as APIs, data, and compute on behalf of humans. The core issue is how to let an agent pay autonomously. Lao Bai links this stage directly to the payments track. He says X402 currently carries some signs that make the thesis look weaker, which is part of why @hosseeb publicly walked back his earlier call. Even so, he adds that the sector is still early and that it is too soon to declare the idea unworkable.
Stage two: agents buying from agents
Once agents begin transacting with other agents, the problem changes. Payment is no longer the only concern; performance and delivery become central. The article says the original piece proposes using bonds or collateral to constrain behavior. Because agent identities can be created at near-zero cost, a reputation system can be gamed easily. Collateral that can actually be seized is presented as a more credible form of enforcement.
Lao Bai says this second stage lines up with the reputation track and points to ERC8004 as an example of work in that direction. Still, he argues that this area, much like X402, also looks weak based on current data. His simplified version of the critique is direct: an on-chain identity does not equal credible reputation; a reputation record does not equal real transaction history; and the cost for an agent to manufacture reputation is low.
Stage three: Disposable Markets
The most imaginative part of the framework, in Lao Bai's view, is the third stage, described as Disposable Markets.
He gives an example from the article: an agent temporarily needs 1,000 hours of GPU compute to complete a task but worries that the supplier may fail to deliver on time. Instead of relying only on a bilateral agreement, the agent could create a market on demand around one specific risk — whether the supplier will deliver on schedule. Other participants could then price that risk and commit capital to take the other side.
Such a market would not need to be listed in advance, and it would not need prearranged market makers. If there is real demand, the system could try to find capital willing to bear that risk. In that framing, the financial market itself becomes infrastructure that can be called when needed.
Beyond payments
Lao Bai's reading is that the first two stages correspond to payments and reputation, but both tracks still face real doubts. The third stage points to something different: agents creating one-off markets for specific risks and then sourcing capital from participants willing to absorb them.
He argues that this is logically plausible in a world of silicon-based actors because their methods of searching for information, discovering opportunities, making decisions, communicating, and pricing differ from those of human society and may operate at much higher efficiency. He is also explicit about the limits of the claim. No one knows whether this will actually happen. For now, it remains an extrapolation, not a verified reality.
The column closes by saying that the launch of Grok Bot and Meta Muse could accelerate the evolution of the Agent Economy. If that happens, Lao Bai expects the outcome to involve more than payment rails alone and to include new forms of financial markets as well.
PANews notes at the end of the article that the piece reflects the views of a contributing columnist, not the outlet's position, and does not constitute investment advice.