The frontier of intelligence, hardware, and the way forward.

First, AI can be frustrating. And a lot of fun. Try squeezing it onto a phone, and the Terminator scenarios quickly give way to a more immediate concern: can we get this thing to finish a sentence? It takes a surprising number of tricks to make a small model work reliably on an everyday device. There’s still a big gap between that experience and what frontier models can do with a data center behind them. Even measuring the gap is tricky.

While building SLIDA, I ran into a wonderfully awkward design problem. I love simple interfaces with a little magic: everything you need up front, easter eggs tucked away, and more to discover when you go looking. But I couldn’t find a simple way to answer a basic question: how capable is this software, right now, on the device in my hand?

With AI, the same app can feel like a very different machine depending on the hardware and models available to it. A familiar interface doesn’t tell you whether there’s a small local model or a roomful of computers doing the thinking.

That’s where the SLIDA meter comes in. By testing the tasks our software actually needs to do, we can give users a simple guide to where their current setup sits: from basic functionality to frontier capability. And we can keep updating that guide as hardware improves and we find ways to fit more intelligence into smaller, more everyday devices.