<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Theorem-Provers on SPERIXLABS</title><link>https://sperixlabs.org/tags/theorem-provers/</link><description>Recent content in Theorem-Provers on SPERIXLABS</description><generator>Hugo -- gohugo.io</generator><language>en-us</language><copyright>SPERIXLABS</copyright><lastBuildDate>Wed, 23 Sep 2026 18:00:00 +0000</lastBuildDate><atom:link href="https://sperixlabs.org/tags/theorem-provers/index.xml" rel="self" type="application/rss+xml"/><item><title>Agents Can't Bluff a Kernel: Teaching AI to Prove Gauss in Five Proof Systems</title><link>https://sperixlabs.org/post/2026/09/agents-cant-bluff-a-kernel-teaching-ai-to-prove-gauss-in-five-proof-systems/</link><pubDate>Wed, 23 Sep 2026 18:00:00 +0000</pubDate><guid>https://sperixlabs.org/post/2026/09/agents-cant-bluff-a-kernel-teaching-ai-to-prove-gauss-in-five-proof-systems/</guid><description>The factorial function type-checked on the first try. It passed the termination checker. It compiled to C, ran natively, and printed an answer. The answer was wrong.
Not wrong the way LLM output is usually wrong — no hallucinated citation, no subtly plausible prose. Wrong the way only a machine can be wrong: fact(10) returned 986410 instead of 3628800, because the agent writing it had expressed the step as 1 + p * fact(p) instead of (1 + p) * fact(p). Every guarantee the language promised, it kept. Types: fine. Termination: fine. Intent: never checked, because intent is not a type.</description></item></channel></rss>