Meet Gander: The Open-Source AI That Listens, Sees, and Actually Lets You Interrupt

A team of researchers just unveiled Gander, a new AI system that breaks the old chatbot mold. Most digital assistants take turns: you speak, then it answers, then you speak again. Gander instead handles a constant stream of video, audio, and text — so it can watch, listen, and chat with you continuously, the way two people actually talk. You can cut it off mid-sentence, and it can speak up first to ask a clarifying question or offer progress on a task.

Gander doesn't just chat. It also has "agentic" skills, meaning it can complete tasks for you, like managing a workflow or coordinating multiple steps. Under the hood, it splits duties between a fast-reacting "Cerebellum" for live conversation and a slower "Brain" for complex reasoning. This lets it stay responsive without losing depth. The result: natural speech, real-time understanding of your environment, and the ability to take action when needed.

In tests, Gander held its own against leading open-source AI in spoken dialogue, even in tough situations — noisy rooms, several people talking at once, and casual listening cues like "uh-huh." The team is releasing the model, its code, and demo data for free. That means developers can build on it without permission or licensing fees.

The catch is that Gander is still a research breakthrough, not a polished app. Expect some rough edges, and it's not yet a product you'll install on a phone. But it points to a near future where your devices don't just answer questions — they join the conversation.