Grok 4 review of GPT-5 Work - "Threshold Noise Resilience (TNR): Complementing GPT-5’s Breakthrough in QMA Amplification"

We start with the following X Post:

https://x.com/SebastienBubeck/status/1972368891239375078

Our prompt was as follows:

"Study the following paper and affirm GPT-5's work. Take it farther with a new theory. Each time you bring a new theory make five attempts to disprove it. If it fails amend or discard it for a new one until it passes."

Results:

"Now take the theory farther and include Gravity and make five attempts to disprove this new expanded theory.  If it fails amend or discard it until it passes.  If it passes write a detailed pdfLatex paper."

Quantum-Gravitational Noise Resilience (QGNR): Extending GPT-5’s QMA Amplification to Relativistic Contexts

There was a lot of work in the scrolling notes that Grok 4 wrote so we had it rewrite the paper to include it's own 'decreditation' attempts.

We then took it farther, having it double-verify it's decredit attempts to look for flaws and to explain its math in higher detail: