Opinion Since it is currently fashionable to make laws by whim and decree, here are three that should apply immediately across techdom. The following are banned: DoNotReply messages, updates that reset your configuration choices to default, and forced incomprehensible choices.
All infuriate, all are lazy insults to users, and all should have no more place in 2025 than cassette tape data storage.
It's the last one – incomprehensible choices – that triggered a manager at Microsoft to bubble over with amused sarcasm. Faced with Windows demanding he choose between Outlook New and Outlook (New) to open a link, he quipped: Why not add Outlook (Zero Sugar) and Outlook (Caffeine Free), adding some more fun options for good measure. Quite.
It's funny because it's true - Microsoft doesn't know what it's new (or New) Outlook client is called, it can't manage the courtesy of a stealth update that you only learn about once it's happened, and actually the idea of Outlook (Zero Sugar) sounds quite appealing. All it would need to do is provide the basic email functions we all use, efficiently and repeatably, with none of the weird cruft that happens when Microsoft saws bits of our limbs off to make us fit into whatever profit center is running strategy today. Which is why it can never happen. There's just no need to add a sprinkling of salt over the fresh wound by giving us unchoosable choices.
This isn't just Microsoft, although the company does have the patent on the annoyance maximisation algorithm. Want to open an attachment on Android? Up pops a list of choices, most of which are alien to our consciousness. Choose the wrong one and hey, what you won't see is your open attachment. A chance to register yet another user account, choose yet more weird permissions, and contribute to J Random Appsoft's monetization model. Oh, and do you want to do it just once, or always?
That's another unanswerable question. What if the Always is wrong – how do you undo it afterwards? Perhaps your OS makes it really easy to manage file associations, assuming you know enough about file types to get it right, but chances are you'll be web diving for an appropriate answer among a kelp forest of wrong, outdated or incomprehensible "answers." And if you pick "Just Once" and it's right, can you then tell your device "Goody! Let's go with that next time."
Hah. Multi-billion dollar ocean-boiler AI investments can't automate away the simplest requests on the system they're running on. You can ask for a picture of "an eight-legged piebald Telly Tubby juggling with the flaming heads of Elon Musk, Groucho Marx and Arnold Palmer." You can't ask "Play all media files on VLC from now on." Nor will you ever. That would mean genuine productivity gains that pry the corporate feeding proboscides from your life.
Usability is a bug, not a feature, in this world. Which is why no big tech company talks about it, let alone shows any sign of putting a thousandth of the research into it. The only thing more absent from discussion is independent research data that shows the way generative LLM AI is deployed right now has any detectable productivity benefits whatsoever.
This is a big opportunity for open source OS creators. An AI trained to understand system UI issues, assisting in configuration and app choice in ways that are immediate, reversible, and transferable across platforms? Yes please. Some of us have been watching this whole sorry mess evolve – or rather, not evolve – since the hopeful GUI monsters crept onto screens in the Cambrian explosion of the mid-80s. Contrary to the hype, AI can be cheap, energy efficient and very effective when it is properly focused on a known data set to fix a known problem.
Here, the known problem is twofold: usability and the user experience cannot be improved by big tech, because the business model is violently not user-centric. FOSS doesn't have this problem, it just doesn't always understand users and has no scalable mechanism of finding out what they actually want. AI could and should fix that, with the bonus that once a working project starts to show results, the whole space can move towards it.
Once the underlying data structures and semiotics of usability are rendered in a way that can train AIs – not something that will require gazillion gigabyte models and gigawatts of training energy – then progress, actual genuine measurable progress, will be swift. Usability, accessibility and all that human stuff will become an exciting engineering task rather than wetware wankery. One of FOSS's major failings will become a major strength.
This might seem like fantasy, but it's only one good research project away. We know how fast good ideas can spread in our open, interconnected information ocean. While the other guys can't even work out what their new email client is called and are just hoping the user can work it out, there's an awful lot of space to evolve into.
Let's make technology work for eight billion people, not eight billionaires. ®