THE YEAR OF MULTI-MODAL

Summary from the TLDR Founders newsletter

Text was the fastest path to usefulness for AI, but it was always a middle state. Humans don't experience the world in text. Work doesn't happen in text. What's changing now is that multi-modal models are good enough, inference costs are low enough for always-on perception, and cameras, microphones, and sensors are making the physical world legible to machines. Once AI can see, hear, and react continuously, text-only AI will feel as constrained as desktop-only software felt once mobile took over.