Multimodal.

Multimodal just means an LLM with eyes. Before, AI mostly worked on text. Now it can actually see.

Instead of only reading words, the model can take in screenshots, video, screen recordings, dashboards, PDFs, and audio, and reason about them the way it does with text. You can hand it a picture of your dashboard, ask what's wrong, and it gets it.

For building, that's a big deal. A model that can see can work with the visual world, not just describe it. It can read an interface, follow what's on a screen, and act on what it sees. That's what makes something like Computer use possible.

If you've been treating AI as a text box, multimodal is what lets you feed it almost anything alongside your prompt.