Point and Yap
I think the most interesting part is that they built it on MCP. That's the smart move. Rather than locking the interaction into one vendor's agent, it plugs into an open protocol, which means this could work across tools and platforms. That's how you get adoption, and it's the part I'd want to dig into more.
I'm a little skeptical about how well the pointing actually translates. On a laptop, pointing usually means a cursor or a click, but on a phone or a big display it gets fuzzier. And voice agents still stumble on context, so the question is whether the pointing disambiguates enough to make the yapping reliable. I'd want to see it handle a messy real-world task, something beyond a clean demo.
Still, this is the direction I keep waiting for. Typing prompts is a temporary phase, and interfaces that let you gesture and talk are closer to how people actually work. If the execution holds up, this is the kind of thing that makes agents feel less like chatbots and more like coworkers.