1. AI is the triumph of capital
If you know your classical economics, you will know production involves three factors: Land (the natural resources of the planet); Labour (the work that humans put in); and Capital (the means of production, from stone tools to factories). We live in the age of capitalism, where capitalists — the owners of the means of production — dominate both economically and politically.
AI is capital. It's a means of production — and as with other new technologies in history, it replaces labour. This article takes a great look at how AI will replace labour, and what the consequences will be. It's not pretty.
2. Bluesky and anti-moderation
Bluesky has seen an enormous spurt of growth of late, and quite a few early users have been complaining. The complaints are mainly that the service has grown, attracted some of the more shouty and destructive people from Twitter, and so has become less of a welcoming space, particularly for some minorities.
Yes, that means that the Nazis have arrived. And the transphobes. And the homophobes. We're not quite yet at the "flooded by Trump supporters" stage, but it can't be too far off.
So what does Bluesky do about it? Well, not that much. That's because the system was designed from the ground up to base its moderation systems on something called "composable moderation", with user-created block lists rather than centralised moderation systems.
There's only one problem: it's a system that doesn't really work. Or rather it works, but it's incredibly easy to use as a tool for abuse:
If someone puts you on a list called Evil People Who Kick Puppies, well, you're out of luck. Now everyone who sees that list and goes "oh man, I hate evil people who kick puppies; let's subscribe to this mass block list" will have you blocked, and there's nothing you can do about it. People you want to follow and people who were already following you can both be ripped away from you. With a mechanism like this, ordinary users are more empowered to wreck your experience of the whole site.
3. Do what we say, but DEFINITELY don't do what we do
Various people have spotted that Apple has implemented a feature called Enhanced Visual Search in recent versions of iOS. On the face of it, this sounds fairly anodyne:
Enhanced Visual Search in Photos allows you to search for photos using landmarks or points of interest. Your device privately matches places in your photos to a global index Apple maintains on our servers. We apply homomorphic encryption and differential privacy, and use an OHTTP relay that hides IP address. This prevents Apple from learning about the information in your photos.
That's all a bit cryptic. Is this matching solely on visual data, or on photo metadata? It's not clear. It's not, in fact, clear what this feature actually does.
Apple says you can trust it, and I don't doubt its claims about how it ensures it can't access your data. What I think it galling, though, is this was enabled by default, and the switch to turn it off buried deep in preferences.
Imagine if a third-party developer pulled the same stunt. No matter how good that third party's encryption, regardless of their promises, Apple would make them ask for upfront consent — and rightly so. But Apple gets to just shrug, say "trust us", and move on.
4. The biggest threat to the independence of judges
Every year the Chief Justice of the US Supreme Court issues a "Report on the Federal Judiciary" and this year it includes an essay on threats to the independence of judges. These include violence, intimidation, and disinformation. But one that's missing: the partisan politicisation of the court itself. When justices and judges and chosen not for their judicial excellence but because of their ideology, the system can no longer be considered independent.
5. Alas, poor Windows 11
This is a very long and detailed look at every single thing which is wrong with Window 11 from a user interface perspective. It's a great read, partly because, if you have used Win11 for any period of time, you will find yourself nodding furiously in recognition. But also it's on Pastebin. Remember Pastebin? I thought that had disappeared…
6. Distraction-free writing (DOS edition)
My old chum Liam Proven has created a bootable USB image which includes a version of DOS plus all the distraction-free writing tools you're ever likely to need. If you want to know what it was like writing before the internet with all its distractions, this is a great way to try it. And you might like it: Protext was a great editor (and was used on the wonderful Amstrad NC100 portable computer, which was basically the poor student's Tandy Model 100).
7. Not so much a link, more a guideline
There’s an old saying, “if you want to change yourself, read more books”. Actually, that’s not an old saying, I just made it up, but I think it nicely sums up something that’s certainly true if you would like to write. You learn to write by reading, and then writing. Just writing isn’t enough.
All of which is a long way to say that I’ve spent the couple of years ramping up the number of books I read, and finding new writers to explore. In 2023, when I started tracking, I read 15 books. In 2024, I read 28. The aim this year to is read at least 35, and as long as I get into the habit, I reckon that could be more.
All of which is an even longer way of saying that I’ve finished the first book of 2025 (to be fair, I started it in 2024, so don’t think it’s all that impressive): Eliza Clark’s She’s Always Hungry.Eliza, like Julia Armfield, is one of those writers that not only makes me go “oooh that’s bloody good” every few pages, but also makes me want to write. Her voice is one that makes me think the stories in my head are worth committing to paper.
If you know Eliza’s stories and are now concerned for my welfare and those around me, please don’t be: I am not either convinced I am an apex predator, nor likely to eat anyone one piece at a time.
And if that last paragraph gives you any kind of strong reaction, I would heartily recommend you go buy this book now.
8. Why does everything connected with wind energy look so beautiful?
Shipping goods by boat is more energy efficient and has less impact on climate change per tonne moved than air freight. But it also accounts for around 3% of global carbon emissions – and that’s estimated to grow to 10% by 2050.
But unlike aircraft, ships can also tap into one of the most bountiful and easy to access sources of energy on the planet: wind. This MIT Technology Review article looks at some of the ways that we can step back to using wind to propel ships. I think these ship designs also look wonderful – which may not be the most important thing, but I think it’s a positive.
9. Tough times for Tesla
You might have noticed that I don’t like Elon Musk much. So you might think that I would be happy that Tesla – still Musk’s main source of wealth – seems to be on the ropes, with declining year over year sales for the first time in its history.
Tesla isn’t the only EV maker that is having issues, though. Mercedes-Benz also saw EV sales drop recently, by a staggering one-third. Does this mean that the west is falling out of love with EVs?
It is of course more complicated than that. While demand for EVs has cooled in Europe and the US, this is partly down to dumb political decisions like phasing out subsidies for EVs replacing fossil fuel engine vehicles, and a lack of mass market cars. Most of what’s out there falls into the “premium” category, which people are reluctant to invest in given uncertainty over the economy.
But it’s not all doom and gloom. Chinese EV makers are experiencing strong sales, and many of the models they make are more affordable – and are likely to hit Western markets over the next couple of years. You can get a pretty good car from BYD for £26,000 list (about £21,000 if you look around), a good £10,000 less than the cheapest Tesla. Price is becoming less of a barrier – but I would expect it to be Chinese manufacturers, not western ones, that clean up.
10. The year in LLMs
Simon Willison has put together a fantastic summary of what’s been happening in the world of large language models this year. The good news: efficiency of models is massively improved, although as Simon notes, there are still environmental concerns over the big build-out of data centres.
I’m still ambivalent about LLMs. I can see the applications and they actually excite me. It’s worth thinking about how the use of them changes as they move from cloud to local. And within a couple of years we will see models more powerful than we can run now being run locally – as Simon notes, you can now run stupidly powerful models on your phone.
The problem, as always, is that instead of thinking of great new things to do with them, our instant focus is on replacing humans. The best use of technology is always to augment what humans can do — the “bicycle for the mind” model – rather than replacing them. I think, in part, that’s because of the framing: LLMs are “artificial intelligence”, which immediately sounds like a replacement for human thinking.
But LLMs don’t think. The latest models are augmented with some reasoning capabilities, but it’s still basic (and boring) compared to what humans can do. So the key question is: How do we use LLMs, day to day, as a tool, not a substitute for people?
Maybe we’ll find out in 2025.