Karpathy Endorses HTML Output for Large Language Models, Predicts Interactive Neural Video as Ultimate Form

According to Andrej Karpathy, OpenAI founding member and “vibe coding” concept creator, today he endorsed the Claude Code team’s approach of using HTML instead of Markdown for large language model outputs. Karpathy outlined an evolution roadmap for AI interaction interfaces: from plain text to Markdown to HTML, followed by multiple intermediate forms, ultimately reaching the final stage of interactive neural video generated directly by diffusion models.

Karpathy attributed this evolution to human brain bandwidth, noting that approximately one-third of the human brain processes visual signals in parallel—a “ten-lane highway” for information input. He argued the optimal human-AI interaction combines efficient voice for human input and high-bandwidth visual output (images, animations, or video) from AI. He recommended users immediately add “structure replies as HTML” to prompts as a near-term improvement.