The marvel of large language models is their coherence. Ask them a question, and they respond with sentences that flow, paragraphs that hold together, and arguments that appear structured. Coherence has been the mark of their success. But coherence is also the sign of their limit.
Why? Because coherence is not the same as truth, or understanding, or thought. It is only the surface quality of language holding together. By scaling models with billions of parameters and trillions of tokens we have perfected coherence to an extraordinary degree. And yet, it is precisely at this perfection that the gap reveals itself. The model can speak endlessly, but it cannot ground what it says. It cannot inhabit a world.
Humans are coherent not because we have seen every possible token sequence, but because we live in situations that make sense. Salience comes from care, urgency, mood, and context. We do not weigh every possible continuation of a sentence, we rather speak from within a world that already discloses what matters. This is what gives coherence its depth, and what no amount of scaling can provide.
So coherence is both the summit and the ceiling of transformers. They can simulate the flow of language but not the ground of meaning. LLMs have gotten as coherent as they will ever get, there isn’t anything more there. To go further requires something beyond scaling, something like World Mind, where coherence is no longer the end but the beginning, layered with simulated structures of world, projection, and care.