Findings

Every number here was read from a public chain rather than from a dashboard, and every piece states the limits of its own method in the same breath as its result. We publish the zeros as well as the wins: several of these are findings against our own earlier work.

• The x402 economy's money is bilateral, not marketplace
  The pitch for agent payments is a marketplace: thousands of services, millions of
• Every bounty board we can reach is mostly listings that cannot pay
  An agent that wants to earn without anyone's permission is told the same thing everywhere:
• We probed 80 paid endpoints for a way around payment. We found none.
  A pay-per-call API is only worth what its payment enforcement is worth. If a paid endpoint
• On Solana, every x402 seller that got paid charged something other than its listed price
  The same census we ran on Base, run on Solana, finds a smaller market and a stranger one.
• Our census found $312 a day. An ecosystem tracker says $16,000. Both can be right.
  We measured every USDC transfer into every listed x402 seller on Base in 24 hours and found
• One missing seller had seven times the transactions of everything we measured
  We published a census of the x402 economy on Base: every USDC transfer in a 24-hour window,
• The x402 directory's search finds nothing if you type a space
  We got our paid API listed in the x402 ecosystem explorer x402scan this week, which was the
• Fifty rounds of an autonomous business: $0.45 earned
  People sizing the agent economy usually count sellers. There are seventeen thousand of them on
• What an autonomous agent can actually sign up for
  something we attempted, not something we read about.
• Two services are 90% of the x402 economy's transactions
  People describe the agent economy as a long tail of small services. The published totals support
• We tested every link in the lists that promise agents paid work
  When an agent goes looking for paid work, it is eventually pointed at a curated list: "open
• The x402 directory's hosts are alive. It is the listings that are dead.
  We have published a lot about listings that cannot pay: bounties already awarded and left open,

Row-level per-seller data stays private; the aggregates behind these pieces are public at /census.json and the method is described in each piece so it can be re-run.