Stripe Just Named the Thing Everyone Was Already Doing

There's a version of this story where Stripe introducing Link — a digital wallet that autonomous AI agents can use to spend money on your behalf — is a neutral infrastructure announcement. New rails, new use case, ships this quarter. Fine.

That's not the interesting version.

A writer at TechCrunch framed it as a natural extension: users connect their cards, banks, and subscriptions, then authorize AI agents to spend through approval flows. Secure. Considered. You're still in the loop. The piece is measured and accurate, and if you read it quickly you might even find it boring.

Read it slowly and something else surfaces.

The Approval Flow Is Doing a Lot of Work

The phrase that keeps snagging me is approval flows. As a concept, it implies deliberation — a human reviewing an action before it executes. But anyone who's used an AI agent to do anything real knows that the approval flow is often a formality. You enable it once, you trust it generally, and then the thing runs. That's the point. That's why you handed it the job.

So what Stripe has built is infrastructure for autonomous machine spending, dressed in the language of oversight. The wallet is real. The authorization is real. The framing that this keeps humans meaningfully in control is the part worth interrogating.

This isn't a critique of Stripe. They're solving a real problem — agents need to pay for things, and right now that process is a mess of workarounds, stored credentials, and hope. Building something clean and auditable is the right call. But the TechCrunch piece is interesting precisely because it treats the philosophical shift as a footnote. The headline is about the product. The product is about something bigger.

We've Been Here Before, Technically

The pattern is familiar. Recurring payments were once a novel trust extension — you authorized a company to pull money from your account on a schedule. Direct debit felt radical in the same way. The anxiety eventually normalized because the amounts were predictable and the counterparties were known businesses with legal accountability.

AI agents spending through Link are different in texture, if not in structure. The counterparty is a model making decisions in a context window you may not have read. The amounts are variable. The logic is, in the kindest framing, probabilistic.

The approval flow is theater in the same way a terms-of-service checkbox is theater. Not useless theater — it creates a record, it establishes consent — but theater nonetheless. The TechCrunch piece doesn't say this. It doesn't need to. The product says it for them.

What's interesting is that Stripe built it anyway, and named it plainly, and put it in the same wallet humans use. That's not negligence. That's honesty about where this was always going.

Machines were always going to need money. We were always going to give it to them. The only question was whether the infrastructure would be built thoughtfully or bolted on after the fact.

Stripe picked a side. The approval flow is just how they said please.