The $236 billion economy that doesn't need your attention

Before I led the Stellar Development Foundation, I spent years at Mozilla fighting the surveillance economy, advocating for privacy and encryption at a time when those positions were still controversial. Those battles were necessary, and some were won. But the $1+ trillion advertising industry adapted every time. GDPR, CCPA, Apple's privacy toggle: none of them changed the default. Behavioral targeting still runs the internet.

I'm not predicting its death. The attention economy and whatever comes next will coexist for a long time. But something structural is shifting, not in the regulations, but in who's actually browsing.

The attention economy grew and a simple bargain took hold: you can't charge people for content, so you charge for their attention instead. Micropayments never took root. Paywalls fractured audiences. Free content subsidized by ads became the norm. I called this the value exchange: we gave our eyeballs, our data, a lot of ourselves, and we got the content, sometimes simplicity, and often task-related ease of use. I was never sure that the value exchanged was equal, sometimes it was intrusive, and often without users really knowing something was exchanged, but it became the economy of the web.

That bargain has now held for decades because of the human in the loop. Someone has to see the ad. Every dollar in digital advertising depends on impressions, pairs of eyes, and the cognitive biases behind them.

AI agents are handling a growing share of tasks that once required human attention: researching purchases, comparing services, managing subscriptions, booking travel. They act on behalf of people, but they don't experience content the way people do. An agent can't be nudged by a countdown timer or a well-placed banner. There's no persistent identity to track. Retargeting fails because you can't follow an agent home.

The ad industry is fighting back. ZeroClick raised $55 million to embed advertising inside AI agent reasoning chains. But the W3C is finalizing WebMCP, a standard where websites declare structured capabilities directly to agents, including what a site offers, what it costs, and how to access it. When agents can discover, price, and pay for services programmatically, the persuasion layer loses ground.

If you can't monetize eyeballs, you need a different economic primitive. The obvious candidate is direct payment, something the internet was supposed to have from the beginning. I wanted this, a tip jar in the upper right hand corner of a browser that allowed a user to tip a content site, send a micropayment, but we could never get the economics right. At that time, the cost of any single payment was too high to justify microtransactions.

But the idea wasn’t wrong. Micropayments failed in 2001 and again in 2015 because rails were built for large transactions, not millions of small ones.

That's changed. Blockchain-based protocols now settle transactions in seconds at fractions of a cent. The HTTP 402 status code, "Payment Required," was reserved in the original web specification in 1997 but never implemented. Until now. 

Protocols like x402 are finally giving it meaning: an agent encounters a paywall, receives a price, evaluates it against its budget and instructions, pays (or doesn’t), and moves on. No ad impression. No data exchange. Value determines all.

And the demand side has arrived. The World Economic Forum projects AI agent commerce at $236 billion by 2034. Major payment networks and infrastructure companies are already building for it. When the primary consumer of web content is a machine acting on a human's behalf, direct payment isn't optional.

It's the only model that makes sense.

Take journalism. The advertising model that was supposed to fund the free press instead created incentives for clickbait and engagement farming. A world where agents pay for information based on accuracy and sourcing would reshape those incentives from the ground up. Or take the creator economy—a payment layer at the protocol level, without a platform intermediary extracting a meaningful share of the revenue, changes who holds power in digital creative work.

But this only works if payment itself doesn't become the new tracking mechanism.

The battle lines are forming. Traditional payment networks are building closed agent payment systems that preserve their intermediary position. Open protocols are building rails where any agent can pay any service—without a platform in the middle, at fractions of a cent. 

Which architecture prevails will determine whether agent commerce reproduces the attention economy's extractive dynamics or breaks from them.

I don't know how fast this transition happens. 

It could be five years or twenty-five. Advertising will target humans in the shrinking number of contexts where humans are directly present. Agent-mediated interactions will migrate to direct payment.

But there's a risk I want to name. Blockchain transactions are readable by default. If what your agent buys, reads, and accesses becomes a trackable onchain profile, we will have replaced the attention economy with an activity economy—agent payments as the new cookie, sellable to the same industry that funded surveillance the first time. 

The tools to prevent this exist today: privacy-preserving protocols, zero-knowledge proofs, transaction shielding. We have a chance to build privacy into this infrastructure from the start. I watched us fail to do that with the last version of the internet. I don't intend to let it happen twice.

When the consumer isn't human, you can't sell attention. You have to sell value. The attention economy won't be defeated by regulation or outrage. 

It will be quietly succeeded by a model that simply doesn't need it.