OpenClaw - The tweet discusses problems with agentic frameworks like

@boom_dart @EricRWeinstein The problem is deeper than that. For example, many agentic frameworks like OpenClaw have state. The problem is how training, tokenization and neural networks produce results in transformer models. There are simply no next token results available to jump beyond training data. Source: https://x.com/FearAndMadnUS/status/2037710130100502666