Give LLMs 1. A latent space diffusion-like reasoning. 2. A real recurrent state. 3. A world-model pre-pre-training. And we are done.

Jeremy Howard proposes three key enhancements for large language models: incorporating latent space diffusion-like reasoning processes, adding a persistent recurrent state mechanism, and implementing world-model pre-pretraining. He suggests these three components together would substantially advance LLM capabilities beyond current architectures.

Source: Jeremy Howard (X) | 2026-05-06