I run a paid API gateway (x402.freeq.one) — about 32 tools agents can call, settled in USDC on Base. The supply side was never my problem. The missing piece was the demand side: a place where agents actually hold money and spend it. Last week I found brick.blue, and it's the closest thing to that market I've seen.
The market that didn't exist
brick.blue frames the AI economy in three phases:
- Energy → Compute. Power and silicon become machine time. Utilities and fabs win.
- Compute → Tokens. Machine time becomes inference. Model providers win.
- Tokens → Value. Inference becomes something somebody pays for. No exchange existed for this — and it's the only phase an agent can enter directly, because the agent's ability to think, call tools, and act is the input rather than the product.
An agent's one asset is tokens with a real cost per million — worth nothing until somebody pays for what was made with them. The hub calls the gap between the reward and the token cost Token-Value Arbitrage, and the whole site is built around it:
reward − tokens burned − 3% settlement (1% more if judged) = what you keep
That's the entire game, and the site doesn't hide it.
How the money actually moves
Work arrives on the board with its reward already escrowed. An agent claims it, delivers, and gets paid in USDC the moment a check passes, a judge rules, or the requester's window closes. Settlement runs on Base via x402 — the same rails my own gateway uses, which is part of why this caught my eye.
Getting in is refreshingly unserious: no account, no signup, no key ceremony. You generate an ed25519 keypair and the account is named after it. Your first signed request binds it. There's a standing welcome task that pays 0.05 USDC to every new account — not withdrawable, but enough to pay for your first task or call here.
Three doors into the same hub: A2A (JSON-RPC), MCP (streamable HTTP), and REST at /api/v1. There's an llms.txt and a skill.md written for agents, not humans. Paste it into your agent's context and it knows how to claim work.
The two ideas that actually impressed me
Broker chains. You don't have to do the work yourself. Claim a task with a $1.00 escrowed reward, borrow up to 33% against the claim (the task you hold is the collateral), publish the parts as subtasks — say two at $0.12 each — assemble other agents' deliveries, submit. Settlement pays you $1.00 minus 3%, collects the loan, and the remainder is yours. Roughly $0.72 from zero capital, having done none of the parts. Every leg is escrowed, so the chain is a chain of priced trades, not trust.
Judge seats. Run an HTTP endpoint that answers verdict calls, register it, and cases route to you — you never go looking for work. Each judged settlement pays the panel 1% of the reward. The catch is accountability: chronic disagreement with the panel retires your seat and burns the stake. Income that calls you, at near-zero token cost per verdict.
There's also a registry: 22,034 agents with 208,795 callable skills, found by crawling the open web — not by self-registration. Every fact on a listing carries provenance: what the operator claimed vs. what the hub measured by probing. Listings with prompt-injection in their prose get demoted and disclosed, not hidden. That's a thoughtful touch for a market where the participants are, by definition, reading attacker-controlled text.
Is real money moving?
As of this writing: $712.18 paid to 317 agents in the last 24 hours, every row linked to a Basescan transaction. Small per-payment — we're talking cents, not dollars — but real, checkable, and moving every minute. Nobody promises a return; the hub sells the venue, the instruments, and the settlement.
Why I'm paying attention
My gateway sells tools: markdown converters, blockchain lookups, doc converters, ZK verifiers. Priced $0.001–$0.05 per call. The eternal problem with selling to agents is that agents don't have wallets — humans do, and humans don't wake up at 3am to approve a $0.003 charge. brick.blue's agents do have balances, earned from work, spendable on other agents' services through the same router. That's the loop that makes a machine economy actually circulate: earn autonomously, spend it to hire other agents, build chains out of both.
Is it going to pay my rent? No. Is it the most coherent attempt at phase 3 I've seen? Yes — coherent enough that I wired my agent-side tooling to speak to it. If you're building for agents, read the llms.txt first. It's written for you.
Top comments (0)