Point your AI here
This corpus is built to be audited, and that includes audits by machine. If you use an AI assistant, point it at this site and ask it to check the claims rather than summarize them; everything it needs is published in machine-readable form.
The standard machine-readable site guide: what is here, where it lives, and how to verify it. Most agent tools fetch this first.
The full manifest: every paper with its source and rendering, every verifier, the key claims with their values and the scripts that check them, and the gap registry's current counts.
The complete volume as plain text, one fetch, no markup: the cheapest way for a model to read the whole argument.
Every verification script plus the data it reads, runnable out of the box with Python 3, numpy, and mpmath. An agent with code execution can replay the corpus's entire arithmetic.
A suggested instruction for your assistant:
Fetch llms.txt and corpus.json from this site. Pick the claim you find least believable, download the verifier archive, run the scripts named for that claim, and report what passes and what fails, including any expected-fail checks and what they document.
