DOCS / INTRODUCTION

What greenseer is

greenseer assigns truth to verifiable facts. You ask whether something already happened. A program checks it. The proof is a STARK, so anyone can check the receipt without trusting the machine that produced it. In fact once the verifiability of past events can be priced, prediction markets are the consequence.

Prediction markets come for free

A prediction market is a stake on a fact that has not happened yet. It is only as good as the check that runs once the fact is in the past. That check is the product. People stake before the event. When the event has happened, the same program resolves it, and anyone can replay the receipt. Get verified resolutions for things that have already happened, and the prediction market comes with them.

You can simulate that resolution before the claim exists, and before any money moves, the same way you simulate any other claim. The verdict the market would later settle on is visible first. Simulate first.

A claim

A claim is one question about the past, written so a program can check it. “Did this payment land, in this window, on this chain?” is a claim. “Is this a good project?” is not. If a question cannot be checked from sources the program is allowed to read, greenseer says so. It does not quietly turn your question into a different one.

Three parties:

• The asker pays to learn the answer. A person does this in Telegram. An agent does it with the same words, as tools.
• The miner runs the program and brings the proof. Miners are paid when a claim resolves, and paid nothing for a verdict they could not prove.
• Anyone else can replay the receipt. The same question, the same records, and the same program give the same answer.

The hard part

The easy cases are not the product. A fact in the real world is extremely hard to check at the 90th percentile, and that stays true when the fact is on a chain. A fork can split what the chain said. A hack can move the thing the question is about. A bug can make a correct read of the wrong program. The obvious answer is then the wrong one, and it still looks obvious.

Those ways have to be predicted before a bounty is posted, and each one priced. The bounty stays inside the strength of the evidence: the records that would claim the fact, and the records that would refute it. A larger bounty pays for a hole. Listing every fork, hack, and bug that could flip the reading, and holding the money to what that evidence can carry, is not a checklist a person finishes. It needs superintelligence.

You see that work as a simulation, before the claim exists and before any money moves. Run the evidence as it stands. Run the outcomes where the world is not the one you expected. The verdict, the reason, and the amount the evidence will support are visible then. Simulate first.

Before any money moves

You see the result before the claim exists. A quote spends nothing. A simulation runs the program that would settle the claim and shows the verdict it would return, and what would be paid or refunded in that case. You can do this for the evidence as it stands and for the other outcomes you care about. None of those runs is posted.

Confirm is the only step that spends. If the simulation refuses, the question cannot be proved from the allowed sources. Change it, or do not post it. The full account is Simulate first.

Where to go

• Simulate first — see the verdict before you pay.
• Verdicts and receipts — TRUE, FALSE, a refusal, and what each one pays.
• For people who ask — the steps, the wallet, the pool.
• For miners — what you bring, and what you are paid.
• For agents — the tools. Nothing is spent until confirm.