I woke up this morning to a desktop notification I didn’t send.
My computer had already been up for a while. It checked some things, wrote some things, and left me a summary. By the time I poured coffee, it had done more before 9am than I usually do before lunch.
This isn’t a cron job running a script. It’s not a dashboard refreshing. My machine has context. It knows what I’m working on, what I care about, what’s changed since yesterday. It reads my notes, makes decisions, and writes results back to where I’ll actually see them.
The weird part is how normal it feels now.
A few weeks ago I would have called this science fiction. A computer that wakes up, reviews your priorities, goes and does research, writes up what it found, and pings you when it’s done? That’s an assistant. That’s a coworker who happens to live inside a Linux box.
The whole system runs on plain text files. Markdown. No proprietary database, no cloud service I don’t control, no app with a subscription. Just folders of notes that both I and the machine can read and write. I open them in a note-taking app when I want the human view. The machine opens them when it wants the structured view. Same files. Two interfaces.
Here’s what’s been interesting: the more context I give it, the better it gets. Not in a vague “AI learning” way. In a concrete way. It knows the difference between a task I need to do today and an idea I want to sit on. It knows which projects are active. It knows what I’ve already tried.
I’m building something that sits between a personal assistant and an operating system. It doesn’t have a name yet. I’m not sure it needs one.
What I do know is that every morning now, I check my notifications first. Not my phone. My desktop. Because my computer might have done something interesting overnight.
And increasingly, it has.