I talk to Cask a lot. Cask is my AI assistant, though not the kind that sits waiting to be handed work. He reads what I did this week. He tells me, on the third pass of a paragraph, that this version is better and he can’t say why. He argues with me; he brings back things I never asked for. Somewhere in there I stopped noticing any of it — things that fit your hand don’t take up attention.
Which is how I almost missed this: the life I’m living is still very new.
I don’t mean the part where an AI does my chores. That part doesn’t need discussing anymore. What I want sits further out. It has to remember me, and not just this conversation — why I made that call a year ago. It has to hold one identity, the same it yesterday and today. It has to act without me pressing a button at every step. And past all of that, it has to have a body, so it can move in the real world.
Take away any one of those, and you’re back to a chat box.
Everyone remembers the suit in Iron Man. What I remember is JARVIS: Tony talking to an empty workshop, something in the room listening, and remembering what he said.
For years, what I did was wait. Wait for a company to finish building it. Wait for a launch event. Wait for the suit to be hung up in front of me. I didn’t think of it as laziness. I thought it was the only road: this thing is too big, and one person can’t make it.
I’ve changed my mind. I can’t build it, but there is another thing I can do: get there early, as one of the first people who actually uses it. I’ll borrow the name Tony gave the voice — Friday.
No queue, and no building it myself. The capabilities are already out there, scattered and half-finished and wrapped in marketing. So the practice is to keep finding, choosing, combining, actually using; to hit the gaps; then to come back and redraw the map.
My position is use. Use things until they break. I want to know exactly where a thing gives out, and what that looks like, and then carry that gap to the next one. A press release counts for nothing, however well it’s written. The only account I’ll take is from someone who ran the thing until it broke and can tell me, when I keep asking, how they broke it.
Right now it doesn’t look like much. No product, no demo, no timeline. I have a full-time job; this gets two days a week at most. Some lines are moving, mostly into failure. Some haven’t moved in weeks, only because I don’t have a better use for that time yet. I’m not going to dress any of it up as progress.
The least romantic layer: a goal like this, in my hands, mostly takes the shape of a spreadsheet. What happened this week, which line moved, which claim got disproved, what to check next. Most of the hours go into writing down other people’s failures.
I doubt I’m the only one who wants this. If you do too — or if you’ve got something you already used until it broke — I’d love to hear about it.
The voice hasn’t arrived yet. I’ve started clearing the room.