Verifiable AI — Now Live
On the line
Per acknowledged error
Ends
Open to active MiroMind Pro subscribers in the United States (excluding FL, NY, RI, AZ, CO) and Singapore. Pro or Deep mode only. See full Submission Guidelines for scope and eligibility details.
How It Works
1
STEP ONE
A factual error, a fabricated citation, a reasoning chain that breaks. Single-prompt only.
2
STEP TWO
Short form: your prompt, MiroMind's response, what's wrong, and an authoritative source. Two minutes.
3
STEP THREE
Not authoritative: Wikipedia, blogs, social media
The summary above covers the most common cases. A few specific cases — multi-turn conversations, file inputs, code generation, ambiguous prompts — need extra context to evaluate. For the full rules with 30+ real examples of what we accept and reject, read the full Submission Guidelines →
The Conviction
The per-user cap exists to prevent gaming. Thirty acknowledged errors from a single user is a serious contribution. If you're that good at breaking our model, we want to talk to you about more than bounties.
Before You Submit
Word-for-word. A single prompt — not part of a multi-turn conversation.
The full answer it returned, ideally with a shareable link.
In your own words. Quote the wrong part, explain why, give the correct version.
A link or citation that proves the correct answer. See "What counts as authoritative" above.
The Promise
Who can submit
Why only Pro and Deep mode?
How fast do I get paid?
How will I receive the $100?
Can I submit multiple times?
What if I find a lot of errors?
What if my error is "already known"?
What if you can't reproduce my error?
What about taxes?
What happens after 6 months?
Why aren't you running this in my country?
$10 million on the line. Starting today.