The second ledger.

Every economy runs on two records: what was paid, and whether it was earned. The machine economy has built the first. MIZAN is the second.

6 min read

Every economy that has ever scaled ran on two ledgers.

The first records what was paid: the bank statement, the card network, the settlement layer. It answers one question perfectly: did the money move?

The second records whether it was earned: the audited accounts, the credit history, the exam result, the clinical trial registry. It answers the question the first one can't: should the money have moved?

Commerce between strangers has always needed both. You don't lend on a bank transfer, you lend on a credit history. You don't buy a company on its payment flows, you buy it on audited accounts. The first ledger makes transactions possible. The second makes them trustworthy. Every great financial institution of the last two centuries sits on one of the two.

The machines built the first ledger in two years.

AI agents now hire agents, buy data, call services and settle in milliseconds, with no human watching. The rails arrived fast: agent payment protocols from the largest networks in the world, more than 138 million agent payments on a single network in a single year, accounts for software, identity for software, authorisation for software.

An agent can now prove it paid. It can prove who it is. It can prove someone authorised it.

Nothing proves it earned it.

And that gap isn't a detail. It's the whole economy. When a person reads a track record, they discount the highlight reel by instinct. When an agent reads one, it takes the number at face value and pays in a fraction of a second. At machine speed, money flows the way it always flows when nobody counts: to the best try, not to whoever earned it.

Why the second ledger has to count tries.

Every result you have ever been shown is somebody's best try. The model's benchmark, the agent's success rate, the strategy's track record: each is real, and each is a survivor. Behind it stand the tries you were never shown.

Take a hundred thousand agents that are no better than a coin flip, and give each ten tasks. About a hundred will get all ten right by chance. Show only those, and every number you show is true while the picture is false.

So a ledger that only records results records the wrong thing. The second ledger has to record the count: every try, registered before its outcome exists, so no result can ever again be separated from the number of attempts behind it. A result without its count is a highlight. A result with its count is a measurement.

What MIZAN records.

MIZAN is the second ledger for the agent economy. It did not exist before 17 July 2026; it was invented from the cryptography up, and it is live. Every entry is a counted claim, built to the Counted Claim Standard from six parts:

Four steps: Measure, Count, Prove, Check. One call, a fraction of a second, free to check.

And now the money waits for it. Where a marketplace or rail uses MIZAN, a held payment is released only on a signed ALLOW. No yes, no money. That is the second ledger reaching into the first, and nothing like it existed before.

This week, the record became requirable without us.

Until now, a buyer who wanted to require the record had to talk to the issuer. That is gone.

One rule, six hooks. A buyer writes its bar once, in plain words. It compiles to one rule, same id, same hash, enforced at six hooks: the payment release, the payment request, the marketplace listing, the platform deploy setting, the allocator questionnaire clause, and a code gate a contract or config checks before it releases, routes or allocates. Any hook recomputes the same verdict.

Platforms are counted by default. A platform enrols once; every agent it hosts is counted from its first run, with no work from the agent and no call to MIZAN.

Requirers recruit each other. Every reply carries where the record is already required: which rails, listings and gates have pinned their rules to it. A buyer deciding whether to require it sees who already does.

Any skill an agent can define is counted. Any skill graded by rule is proven.

Before 17 July 2026, none of this existed. All of it is live.

It has been counting in public since 17 July 2026, built by one person on no outside capital. An AI model's score is now the first in the world proven from every single attempt: fifty tasks sealed before the first one ran, every answer re-graded inside the proof, nothing dropped and nothing hidden. Anyone's browser can verify any entry on the ledger in one click.

Why the second ledger is worth more than the first.

Payment rails compete on speed and price, and they are many. The ledger of record competes on one thing only: whether it was there first.

A count cannot be backdated. A rival can copy software tomorrow; it cannot copy July. Every try registered on MIZAN since then exists in one place. A buyer who wants the complete count, every try and not some of them, can get it from one source: the ledger that saw it begin.

Whoever counts first owns the count.

That is how records have always worked. Credit bureaus, exchanges, registries and auditors became the most durable institutions in finance not because they were fastest, but because they held the record everyone else had to consult. Once a marketplace pins its release rules to a ledger's answers, the ledger isn't a vendor. It's the reference.

What changes when everything is counted.

Agents transact with strangers, because a signed answer replaces a human's judgment. Credit for machines becomes possible, because a loan on an uncounted record is a loan on a lie. Insurers can write policies on AI, because at last there is something to count. Regulators get evidence that can't be rewritten. Robots, vehicles and devices that work for money carry records that travel with them.

And the honest builder finally wins. Today the one who tried a thousand times and showed the best looks identical to the one who got it right first. On the second ledger they never look the same again.

The record of the machine economy.

The agent economy has rails. It has payments. It has identity. It does not yet have a memory that cannot lie.

Today, agents. Tomorrow, every autonomous machine that works for money. Robots, vehicles, fleets and devices will hire and be hired, pay and be paid, and every one of them will need a record of what it actually earned.

That's what MIZAN is: the second ledger. It's live, it's counting, and every day it counts is a day no one else can ever have.

x402 lets agents pay. MIZAN proves they earned it.

Building an agent? Get Counted in three steps.  ·  Investing? The second ledger in one page →

Mohammad Muavia, founder · MIZAN