ebms

A critical take on the scaling argument for AI reasoning, arguing that autoregressive LLMs cannot achieve correctness through more compute alone, and highlighting alternative architectures like EBMs and formal verification as superior for critical applications.