This is a very interesting approach. I naturally think this relates to LLMs, we could in principle encode a Differential DFA chain as additional inputs for AI m

This is a very interesting approach. I naturally think this relates to LLMs, we could in principle encode a Differential DFA chain as additional inputs for AI models, so they could gain capabilities to run symbolic operations reliably, a Neural CPU.