DOCS / INTRODUCTION / SIMULATE FIRST

Simulate first

Before a claim exists, and before any money moves, you can run it. You see the verdict the program would return, and what that verdict would pay.

Why this comes first

A claim puts a bounty, and sometimes other people’s stakes, on an answer. You should know what the program will do with the question before that happens. Simulation is that look. It uses the same program that would settle the claim. It is not a guess, and it is not a vote.

That same resolution is what a prediction market settles on, once the fact has happened. Simulating it now is how you see the check the market would later trust.

The ways it can go wrong

At the 90th percentile a check is not “read the number”. Even on a chain, the chain may have forked, a contract may have been hacked, or a bug may have made the read describe the wrong program. Each of those is an outcome you can run before the claim exists.

The bounty has to stay inside the strength of the evidence on both sides: what would claim the fact, and what would refute it. If a way the fact can fail is stronger than the amount you meant to post, this is where you see it, and where you post a smaller bounty or do not post at all. Predicting that list, and pricing it, needs superintelligence. The run in front of you is that prediction, done before any money moves.

What you see

You give the question. greenseer shows the claim it becomes — the sentence with the blanks filled — and then the program’s result on the evidence:

• The verdict. TRUE, FALSE, or a refusal. A refusal means the program could not prove either from the sources it is allowed to read.
• The reason. Which reads the verdict rests on, in words, not a score.
• The money. What would be paid, refunded, or unpaid if that verdict were the one that settled a real claim. A refusal pays nobody. An expiry refunds the bounty. TRUE or FALSE pays the miner and the side that was right.

You can run it more than once, on the evidence as it stands and on the other outcomes you care about. The payment landed. It did not. The source was missing. Each run is the program, not a story about the program.

What a simulation is not

• It is not posted. No one else can see it, bet on it, or mine it.
• It spends nothing. The quote spends nothing either. Confirm is the only step that spends.
• It is not a receipt and it is not evidence. When a real claim settles, the thing anyone can re-run is the miner’s receipt. Your simulation does not become that receipt.
• It does not make the odds. If a pool opens later, the stakes in it are beliefs. The program decides.

When it refuses

A refusal before you post is the useful case. The question cannot be proved from the allowed sources, and you have learned that for free. Narrow the question, name the missing fact, or do not post it. Posting a question the program will refuse does not produce a paid “no”. It produces no payment at all.

After you confirm

Confirm posts the claim and spends the bounty. A miner runs the program on the live source and seals every read into a receipt. Anyone replays that receipt and gets the same verdict. That replay is the same kind of check as the simulation you already ran — done on the proof, after the claim exists, by whoever wants to check it. The receipt is a STARK: checking it does not require trusting the machine that produced it.