Why Nova exists.

The single most useful thing an intelligence can do is not forget. Memory isn't a feature we added — it's the premise. Everything else follows from a tool that keeps the thread of your work instead of starting over every time.

Context is expensive to rebuild and cheap to keep. Nova holds the projects, people, files, and decisions that make your work make sense, so you never pay the re-explaining tax twice.

A tool that starts from zero each time can only be as good as your last prompt. One that remembers gets more useful the longer you use it — the work accumulates, and so does the understanding.

Your work should be understood on the machine that holds it. Nova runs on-device by default; the cloud is an option you reach for, not a place your work has to live.

The default is the promise. Your screen and files stay on your machine, memory is encrypted at rest, and nothing is used to train a model. Privacy you have to configure isn't really privacy.

You should be able to see what an intelligence knows about you. Nova's memory is inspectable in plain terms and erasable in one command. A black box you can't open isn't trustworthy, however capable.

Nova plans in small, reversible steps and asks before anything irreversible. It's built to act on your behalf, not instead of you — you stay in the loop, and you keep the undo.

We'd rather show than tell. The strongest claim we can make is a real task, unedited, that you can reproduce yourself — so what you see is what the software does, not what a slide says.

Nova is early, and we say so. We mark what's real, what's in development, and what's still vision — plainly — because a product you can trust starts with a company that tells you the truth about it.