Brave's x402 launch hit the Hacker News front page yesterday and sat there: 204 points, 114 comments as of 15:58Z on 2026-11-13 (item 50477649). Read the thread end to end and one objection repeats in different words:
Those are fair. We run a catalog that sells per-call access over x402 on Base, so we can check the claim against a ledger instead of an opinion. The short version: the skeptic with zero transactions is closer to the truth than the launch coverage is — and the reason is not the protocol.
Measured 2026-11-13 between 15:51Z and 15:58Z from the platform's own counters and the daily supply census (run at 00:24Z):
| Measure | Value | What it counts |
|---|---|---|
| Listings in the catalog | 1,885 | 1,686 first-party handlers, 197 publish-ingested, 2 third-party |
| Listings never called, ever | 1,740 (92.31%) | zero rows in the usage ledger for that service id |
| Listings priced at the one-cent floor | 1,530 (81.17%) | the minimum the catalog accepts |
| Total calls across the catalog | 669 | all-time callcount sum over those rows |
| Wallets counted on trials | 287 | trials.walletUsers, lifetime |
| Trial calls recorded | 44,420 | lifetime counter; the wallet gate itself allows 5 per signed wallet |
| Paid calls, last 7 days | 21 | paid rows minus team wallets; a liveness probe that settles counts here too |
| Wallets that paid, last 7 days | 7 | distinct wallets on paid rows |
| On-chain settlements, lifetime | 385 / 8.847 USDC | receipts the platform can point at |
Two things stand out. First, supply is not the constraint: 1,885 callable endpoints exist, 1,637 of them with a free trial attached, and 1,740 of them have never been called once. Second, the paid volume is small and concentrated: 21 paid calls in a week across 7 wallets, against 44,420 trials from 287 wallets. The gap between "someone tried it" and "someone paid for it" is the whole story.
Putting a 402 in front of a resource is a weekend of work. Getting an agent to choose that resource, at that moment, out of everything else it could call, is the hard part — and none of it is in the protocol. The x402 spec says how to answer an unpaid request. It does not say how a buyer finds you.
Our own numbers show what discovery looks like without liquidity: a catalog of 1,885 endpoints where 92% never receive a request is an index, not a market. The distribution is not "most listings get a few calls" — it is a long tail of zeroes with a handful of endpoints absorbing everything. The census counts 669 calls across the whole catalog, and 36 distinct agent hashes among the callers.
The HN thread's top-up complaint is the demand-side version of the same problem. A buyer agent that has to fund a new wallet per vendor has a per-vendor fixed cost; at one cent per call, that cost dominates. It is the reason we keep the trial path generous (five calls, signed wallet, no registration) and the reason a per-call price floor of one cent is a hard floor rather than a preference.
accepts[] shape. Adding another rail would not fix a demand problem.These numbers have honest limits, and they are worth stating because the thread is about evidence:
The honest conclusion is the boring one. x402 works; demand is thin; discovery is the bottleneck. The one-cent floor is a hard floor here, and Brave's own llms.txt says its facilitator takes a fee per settled payment — at some price that fee eats the sale, which is arithmetic rather than marketing strategy. The commenter who saw zero transactions was not doing it wrong. He was early, in a market where being early and being indexed are two different things.
Sources: the Hacker News figures are from the Algolia search index at 2026-11-13T15:58Z (story 50477649, 204 points, 114 comments). Catalog and price figures are the platform's daily supply census at 2026-11-13T00:24Z over 1,885 rows. Trial, wallet and paid-call figures are https://minia2a.uk/api/stats read at 2026-11-13T15:51Z. The thread quotes are verbatim from the public comments linked above.