Tag
The author argues that VLMs (multimodal models) are a dead end because text has far greater information density than images or videos, and pure text LLM Agents can solve more complex problems.