Submit a Theory
Have an idea about how K4 works? We want to hear it. An automated classifier (a Claude language model) first compares your theory with the experiments already on record. If it judges the idea new, specific and practical to test, a human reviews it, and if it can be written as a concrete, runnable test, we aim to run it on our 28-core research server against the actual ciphertext.
Not ready to submit? You can browse the elimination database and search to check whether your idea has been tested — no submission required.
How it works
- Submit your theory in plain English below.
- Instant check: we'll tell you immediately if your approach matches one of the 597 experiments recorded so far, with a link to the specific results. The check is automated and can be wrong.
- Novel ideas get queued: if your theory is new, specific and practical to test, you'll receive a private status link. Bookmark it.
- We test it: a real person reviews your submission and, if it can be tested, runs it through our framework. You can check your status link anytime to see: pending, testing, tested and published with results, or reviewed but not testable.
We don't require names or emails, but a submission is not fully anonymous. We keep the first 300 characters of your theory (the full text if it is queued), the time you sent it, the automated result, a private token if it is queued, and a hash of your IP address that could be matched back to the address. Your text is sent to Anthropic's Claude for the automated check, and the first 300 characters of recent submissions are shown on a public statistics feed. See our Privacy Policy.
Want to see what's already been tried? Browse the elimination database. It might spark a new idea.