Lonely Robots And Other Internets

Orbital Operations for 1 June 2025

In partnership with

Hello from out here on the Thames Delta, where I’m in the midst of finishing a major project for a major comics publisher that probably won’t even be announced for a couple of years because that’s just the way things are now, and trying to stop myself from falling down reading rabbitholes. As I write this, I have 21 tabs open, including the Rabbit R1 “Rabbithole” AI OS interface. I asked it to create a chatbot that will integrate with WordPress. It did. Apparently. I cannot understand the technical instructions to implement it, which seem to involve Linux, a python, a flask and something called Beautiful Soup. It may have hallucinated the whole thing. Hell, I might have hallucinated the whole thing.

This has been triggered by an old acquaintance of mine offering to send me a spare unit of something called the Bee Pioneer. It’s a device the size of an old Fitbit, worn by strap or clip. And it’s an AI listening machine. It’s intended as a memory device: listening to you, reading your Gmail and contacts and such, recording what you’re saying and doing and sending you actionable summaries. It works with an iOS app.

I already have an AI companion device, the Rabbit R1, and it’s a cranky piece of kit. I figured that would be the start and end of my journey with AI devices. I was amused to see Jony Ive and Sam Altman non-announce a non-device the other day, but not deeply interested. And then Dan said, after the poll last week, hey, want one of these?

But, actually… maybe I do.

Dan Catt himself has an office AI called Kitty. Using code I cannot comprehend because I have never been a coder, he’s created a system that keeps him on task, remembers stuff and sends him prompts. I cannot find a link for this right now, and had to ask Perplexity AI for a rundown - there’s also a video by Dan which I’ll put at the bottom of this.

If I had one of those, I would call it Box.

So, listen: if you always think of AI as “machine learning” and not artificial intelligence, you’ll realise you’re already using it every day. Every time you use a search engine, in fact. You just need to notice it. There’s a narrative afoot that AI is a black box, and nobody really understands how it works once you get past a certain level. That narrative serves other interests. It’s algorithms, people write them, people understand them, it’s just really long and complicated.

Remember that bit in THE BIG SHORT where Michael Burry does the thing nobody else could be bothered to do, which was read all those thousands and thousands of subprime loan policies wrapped into those CDO boxes? Algorithmic auditing is how that works in AI, supposedly. You don’t hear much about it. It’s not sexy. Like paging through a million sheets of arcane financial documentation. But it can be done, and it helps you understand that these systems are not mysterious. It’s just maths and spreadsheets.

Anyway. My digression is digressing. I’ve gone analogue and tactile. Why would a coastal hermit want AI devices? Well, if they can ride along in a pocket or be clipped to a shirt, why wouldn’t I? I write things down to remember them, but if a robot can listen to me and remind me of things, that doesn’t hurt. If I can speak to the robot, tell it to research something for me while I’m working and email me the results, why wouldn’t I? As I write this, the Rabbit R1 is making me a 300-word precis on algo auditing citing three sources.

The Rabbit, with its push to talk button, is a walkie-talkie connected to a lonely robot.

Because: what if assistive technologies could just melt into the background or present as intentional physical tools associated with muscle memory? What if we didn’t use these things as toys, or to have them generate some slop to sell? Or, what if we didn’t listen to the current conversation which wants us to treat them like they’re hired friends and we’re the lonely robots?

Not everyone is going to want some random device reading all their emails and shit, and Dan’s already warned me to keep the Bee in a faraday box when I’m not using it. But I feel like it’s time I properly poked my head over the parapet to see what’s going on out there.

And there’s another thing going on here. Muskie is all in on Grok AI because you have to be on X to use it and he needs to keep people on X. He’s right. This is actually existential for him, because AI is Other Internet.

AI chatbot functions are off the streams. It’s really hard to doomscroll a bot window. The Rabbit doesn’t serve you an ad for every three queries. Perplexity doesn’t follow up a task with “and here’s five posts from some guys who hate women, three Nazis and a dude trying to sell you conflict-free cable ties."

When you’re using an AI assistant, you’re not on social platforms or the general corporate web. The assistant is often reading it for you, which also often means it’s looking at the ads for you. Given that ads are served by bots, it’s bots reading bot ads and no human is involved. My poor robot is having to look at hoodies made for gear sticks and videos of retractable awnings to get me what I asked for. AI tasking detaches you from the general internet and actually keeps you in the physical world. And that’s the nightmare for some people - AI accesses the layer of knowledge under the internet we’ve made, and I can get it to write me a summary of some cultural phenomenon while I' go out for a walk or to water the plants.

I take the box out of my chore jacket, wave it at some weird-looking plant that’s emerged, ask the box what it is, it takes a picture and tells me out loud what the plant is. I put the box back in my pocket. There’s no notifications, no “yes, and.” I go back to work.

I was positive it could not be a raspberry, because I was sure nothing could have self-seeded in that spot. The box confirmed it is in fact an extra raspberry. I win. Raspberries.

What’s interesting to me about all this is that it is, on some level, disconnective. It bypasses the loud, stupid, immiserating internet. What if AI is the thing that just gets us what we need while we’re out living our lives?

What if we understood AI as a drawer of tools we can pick up and put away as needed, that keep us off the internet that makes us all so sad and distracted?

What if: the AI bros who are selling us the fairy stories of what AI might do know they’re bullshitting and are panicking because this is as good as the tools are going to be for a fair while (if not worse, because AI models are being made to ingest AI slop), and the adtech bros who profit off keeping us angry know the day is coming when they’re going to be shit out of luck?

Yes, there are all kinds of ecological and political questions around AI, you’re right. But, as researcher Meredith Whittaker said, being right is not a strategy. This stuff isn’t going away, so it’s down to us to define the world we want and loudly reject the bad and greedy and kleptomaniac parts.

I’m still working out my own position. It’s a fluid thing, as I learn more and decide more, and I’m on nobody’s timetable to find that position. But I tell you now. Techno-hermit with talking robot libraries that do stuff while I’m working in the garden, writing in the notebook, sitting quietly by the sea or hanging out with the cat? That’s not the worst possible outcome.

What do you think? Have I lost the last of my marbles?

Letters about the creative life by Warren Ellis, a writer from England. Was this forwarded to you? Subscribe here for free.

Kody Chamberlain and I did a piece for a new anthology graphic book, LIGHTS OUT, which was crowdfunded, but now you can pre-order the book direct. The book contains a piece by Dean Haspiel that should be an awards contender for next year. More details on the project here.

The Humble Bundle has a few days left to run. Good way to support a good charity and explore some of my backlist, if you’re interested in that.

This is out next month, a full collection of the 12 issues by myself, Bryan, Paul Neary and Laura DePuy Martin, in the “Compact” 6×9 format. Which actually works pretty well. It feels nice in the hand, portable but with heft, and it’s going for something like ten bucks, I believe. It’s weird to see it all again.

Also out this month, this $150 paving slab containing several hundred pages of some of my earliest work for Marvel Comics.

GOT MORE TIME?

LTD

I keep a digital writer’s notebook and you’re invited to read over my shoulder.

Morning Computer: a few useful things first thing in my day

Nine Bells: evening notes

HOB’S LANE: new this week, parts 11 - 15

This letter has been zapped to you via Beehiiv and is sponsored by:

The portfolio that's automatically up to date with your work.

  • Authory saves you hours with a portfolio that's always up to date.

  • Get backups of all your articles.

  • Be ready to impress potential clients and employers, anytime.

Well, that was supposed to be a 400 word note because I’m busy right now. Shit. Please don’t unsubscribe, I’ll be good next week and won’t drop an entire non-fiction book of my random and generally useless notions about AI. We’ll talk about, I dunno, trees and music and how I still want to murder my family for giving me an antique ear trumpet one Christmas. You take care of yourself, and I’ll be back next week.

Wasn’t actually joking about that, no.

I’m represented by Angela Cheng Caplan at the Cheng Caplan Company and David Hale Smith at Inkwell Management. Please add

to your email system’s address book or contacts.