A reminder that error messages need to be actionable by the user. Even if that action is "there's a problem at our end, please stop trying for a little while, this isn't going to work right now." Infinite spinning rainbows are not sufficient. Humans have other things to do - give them that hint.

Well, that's probably this, then.

Meta - metaverse-averse.

I am enjoying every part of Belfast except the food, which has so far ranged from acceptable to possibly a hate crime.

One of the solutions proposed - I am not kidding - is "writing scripts to automate repetitive tasks." It's really funny imagining a software engineer being like "woah...like automating the boring stuff, you might say?"

Pope Leo this week called universal health coverage a "moral imperative." He also said: “Health cannot be a luxury for the few,” it is “an essential condition for social peace.”

I'm real pleased that you're implementing the feature where the computer helps serve the human's needs, but you could have done that at any time without the AI. Still can. But now you'll be draining multiple lakes to do so.

Sure, OneDrive status page, everything's fine with the service. Meanwhile, I've been here waiting for a signin for hours, and through a couple of reboots.

I've definitely thought how interesting it would be to have a whole realm of computer science built around ternary logic (whether you view that as "definitely true", "definitely false" and "inconclusive" or something else), but didn't fancy being an impoverished academic to pursue it.

At what point does someone decide that they could just run their economy on legitimate IT services?

Windows Notepad supports Markdown. Windows Notepad Printing does not support Markdown.

Love that Microsoft emails me threatening to turn off my Azure account because a $0.30 charge bounced on my credit card (the credit card company has decided to start treating numerous small charges as suspicious). Clearly, they're hurting for AI money if they scrabble for 30 cents behind the couch.

Those who 'circle back' and 'synergize' also tend to be crap at their jobs https://go.theregist...

It's always fun to notice how the same people who say we "have to be in the office to protect people early in their careers who need to learn" are also saying "we can replace all our entry-level hiring for the next decade with AI".

As an owner of a 3D printer, I have found at least a half-dozen different and unexpected ways to grind gears.

There's an irony to the fact that the bulk of what LLM(AI) does that makes it useful is collecting and distilling the information that search engines should already have been presenting to us, but which they were prevented from doing by paid results and SEO (search engine optimisation).

Is he going to demand to be called "Howard Thunder" now?

Not much of a dad makes much of not much of a dad joke.

I often come across as anti-AI. I'm not. I'm against people using specific forms of AI in ways they don't understand and can't evaluate the risks of. Particularly LLMs and other apparent thinking machines. They don't think, they use probabilistic modeling to derive and depict expressions of thought.

I don't laugh at those who lose a bunch of money in cryptocurrency, but I can understand people who do. My perspective - you already lost that money when you put it in; you handed that money to someone taking you for an idiot. You had the chance to get it back, by taking someone for a bigger idiot.

Every now and then, I'm reminded that politicians can astound and delight me by doing their job in a way that considers future implications on real people. The last time I saw that was with the Data Protection Act in the 80s, so it's about time for another one.

I don't know what's screwed up with Squarespace's payment system - first, they trigger a fraud alert when trying to renew, then when my card company "sorts that out" and gives me approval to charge again, they trigger another fraud alert that kills my card, and now they expired my domain.

You can cut costs further by slicing your bagel in a moebius strip, so you only have one surface, not two, for spreading the cream cheese on.

Also, when you hear "Science can't explain why...", it's either an unrepeatable event with unreliable witnesses (in which case it's open to several interpretations), or there are several interpretations and there's debate over which carries the most weight.

Trend prediction: Traitors UK Series 5: Oops, All Cops.

Some day, science will find a way to breed a seedless everything bagel. And then patent it, so they can sue anyone who grows their own.

Dangerously close to my trademark "every-other-thing bagel".

The Amazon Dash Button was a real thing. I did not dream it.

Sure, you could just ban the word "Microslop" from your Discord, but maybe the AI can help you figure out what the problem is...

"Flush it out" and "flesh it out" are two very different things to do to an idea. They are not the same.

Callers to Washington state’s driver’s license agency who select automated service in Spanish are instead hearing an AI voice speaking English with a strong Spanish accent.

We had a support technician who closed almost all of the tickets she was working on within seconds of receiving the call, too. She'd tell the customer "rebuild database indexes, and call back if that doesn't fix the issue", then hang up and mark the ticket as successfully closed.

Pub's progress... 2008 - The Elm Tree 2012 - Duke of York 2016 - The Duke March 2025 - The Elm Tree again. They're celebrating their 1st birthday under the new name, and new management, in a couple of weeks.

I heard a new, but not surprising twist on the abusive "phishing tests" front today - using AI to impersonate an executive's voice to cajole, bully and outright threaten an employee to give credentials out. Can we stop with this thing of abusing employees in the name of security?

Ridiculous levels of oversight naturally lead to ridiculous levels of evasion to satisfy the oversight, rather than actually getting the work done, evasion of which can be seen by, you know, monitoring the actual work.

Where's the "Block Thread" button? I want to not only hide this sort of claptrap, I want them to know I'm hiding it, and anything else like it, because I want them to shut up about people's delusions about AI. This person needs help and/or silence, not press coverage.

Self-licking ice-cream cone company complains about other company licking the ice-cream they already licked off someone else's ice-cream cone.

I'm pretty certain that the goal isn't to write COBOL fast, but to do it correctly. But you know, whatever floats your stock price, I guess.

CEO of self-licking ice-cream cone company declares self-licking ice-cream cones are cheaper than having humans lick their own ice-cream.

This just means in three years, it'll have eaten senior jobs, in five years, leads, etc. Stop killing the pipeline's entry point and complaining about the flow.

Using software these days it's hard to escape the thought that a lot of the effort that's currently put into pursuing AI should instead be put into improving UI.

The LLM isn't thinking, any more than the submarine is holding its breath.

I feel like there should be an Oxford Book of Victorian Diss, where you could find such gems as: "It was very good of God to let Carlyle and Mrs Carlyle marry one another, and so make only two people miserable and not four."

Surprised it wasn't the orcas wot dunnit.

One of the Old School hackers is gone. Jason was smart and kind, one of the good ones — not just because of his skills, but I have always regarded him fondly because he saw me as a fellow hacker at a time many of his peers could not. I will miss him. Rest in peace, par. 🕊️

If Derivative AI can model an interaction that you believe is indistinguishable from a human, this is not an indication the AI is alive, conscious, or human. It's an indication that most of the humans you meet either are boring, or simply express themselves in a boring way around you.

Amusing outage at work so bizarre that I don't even have a theory as to how it was achieved.

Pro tip - when trying to figure out what the optional arguments for the Linux "head" command are, think very carefully about what you're going to get back when you search the web for "man head". Don't even think about getting help on the 'tail' command.

I have a suspicion that a lot of people are going to be buying 1TB storage cards and find out that they're only equipped for 64GB, but report that they can store 1TB. Be careful out there, and test your storage is capable of actual max capacity.