David M comments on How would a language model become goal-directed?

[David M]

OP here. Now that it’s 2026, it now seems blindingly obvious how LLMs can be repurposed as agents.