Hi friends, two really random and unrelated updates for you:

• Last week I gave a talk about platform decay at Newspeak House in London. It was great fun and I hope to do more — and next time I will tell you before it happens so you can actually maybe come, if you want. Above you can see me chugging a beer in front of the best slide. Thank you to the event organiser andyfor both having me on to speak and also managing to take this very perfect photograph that will now go at the top of my Hinge profile.

• For my girlfriend’s birthday, I concocted an email series that emails you a paragraph about a different pasta shape every day for 28 days. It’s 5% informative and 95% nonsense (with a 100% satisfaction guarantee). She really enjoyed it so I thought I’d share it with others seeing as the automation is all set up and ready to go. If you like my writing, you will love these emails. Get them by going here.

What is hyperpersonalisation?

Recently I’ve been really obsessed with watching old Marvin Minsky lectures. For one thing, his voice is incredibly relaxing and is one of the most effective sleep aids I have at the moment. For another, his large wet brain appears to contain boundless knowledge on computer science and AI. He was making pretty sharp insights about what we now call generative AI all the back in the early 2010s (and maybe even before that).

For example, in this lecture, he noted the difference between engineering and making art:

“Well, when you do a painting, it seems to me, if you're already good at painting, then 9/10ths of the problem is, what should I paint? So you can think of an artist as 10% skill and 90% trying to figure out what the problem is to solve.”

Whereas, he points out, an engineer spends only 10% of their time figuring out what problem to solve, and the rest of the time solving it. Now, if we accept that the image outputs of generative AI systems are ‘good enough’, then we are all already ‘good at painting’ — and therefore we can actually devote 100% of the time to ‘figuring out which problem to solve’. Which, with generative AI, is just deciding on the contents of your image(s).

So basically, depending on how you look at it, we now have ‘infinite problems to solve’ because the only limit is your imagination how good you are at writing prompts. But, once the prompt barrier is out of the way — as I’ve mentioned before, prompt engineering is exhausting and awkward — all we’ll be doing is deciding what content we want to see. The other day I literally asked ChatGPT to show me an image of a thousand people taking out the trash at the same time in suburban America. For no reason other than I thought it would be funny.

This is hyperpersonalisation. I speak to the content machine and it provides. This content is created for no one’s pleasure but my own. I am only showing this to you to make a point. I don’t think this image is interesting to any one but me (and even for me it’s barely interesting). I am not so proud of it that I want to show it off, because I barely did anything to make it exist, besides think of the concept and then type it into a box.

So, ‘personalisation’ in the traditional internet sense is now starting to skew into a different amalgam of experiences. It feels like we’ve gone from be able to change the background colour of our MySpace pages, to receiving ads that we are perhaps ever so slightly more likely to click on, to where ever we are now; where everyone’s social media feeds are so catered to their tastes, and so distinct from each other, that it’s beginning to feel like ‘shared online experiences’ don’t even exist any more — your just one person in a sea of others commenting on a TikTok video.

But hyperpersonalisation takes us one step further from this — monetising collective attention (in a traditional social media feed full of user-generated content) might seem pointless if everyone can just tell an AI model what they want to see. Every social media app already asks you about what your interests are when you sign up; an LLM and image generator wrapped in a decent UI could easily do this too, and synthesise content for your individual pleasure. Imagine, rather than checking boxes in a dumbass app (telling it that you like ‘sports’ and ‘healthy eating’ and ‘video games’), you can go into insane granular detail because you don’t have the generic interests of the collective internet holding you back. There’s no space for mass appeal anymore; just sit back and watch the cast of Succession compete with each other in a street dancing contest, while dressed as furries. Or an alternate timeline of Friends where Rachel and Monica confess their love for each other, everyone cancels Ross, and Phoebe stars a podcast about holistic farming.

I’m not saying that this is where we’re headed, but our particular dark and gritty timeline certainly took a turn when, for instance, Uber appeared and said ‘you can have your own personal driver’ for very little money. And now, thanks to Apple, you can wear a personal computer on your face (for… a lot of money). Can you think of a better way to cut yourself off from your fellow humans than sanctioning your own eyes behind a heavy set of electronic goggles? Here’s a deranged man failing to watch a Mr Beast video while riding the train, while everyone around him succeeds at watching videos on the train because they are just using their phones.

Right anyway, just to be clear I don’t think we’re about to witness the disappearance of content that is made for the sole purpose of being shared far and wide. Rather, it’s that it’s never been more easy to make content for the sole purpose of consuming it yourself — and sharing it is secondary.

I think I had more points to make but it’s been a tough week so I’ll be leaving it there. Next time I think I’ll be sharing some thoughts on Arc Search. Have a great weekend!