OK, I have two questions

I was in a taxi the other day, an electric Toyota, and it was constantly bugging the driver with various “driver safety aids”, like literally a stream of bongs and chimes and sirens. And half of them were “driver inattention! Sit up and look forwards!” He was sitting up and looking forwards, but also he was black, so the car’s interior camera was not registering him properly. But hey, tech is neutral and AI doesn’t encode bias.

“At first they didn’t come for me, which I found remarkably friendly and inclusive.”

People like Elon Musk and Peter Thiel are not geniuses. They're just evil morons who have been able to take advantage of a system built by and for evil morons. We need to rebuild a society where these idiots cannot get money and power.

MacOS 27 defaults* to generating AI slop from your Pictures folder and vomiting it up as your desktop wallpaper. Fuck ALL the way off with that bullshit 🤮

To disable: System Settings ➡️ Wallpaper ➡️ Scroll to bottom where you’ll find a new collection anonymously called “Pictures” ➡️ Click the teeny tiny x to remove it ➡️ Scroll back up to the photo or collection you actually want ➡️ Click that

*possibly only if you previously had it set to shuffle a folder of images

Well, because of some changes in my life, it's time to retire the photographic avatar and go with my old friend the AG Systems mascot. My apologies to @girlonthenet in particular.

“There are decades when nothing happens, and there are weeks when decades happen.” And after a week when I have got the grades to continue my MA, said a sad final goodbye to my beloved cat Nishi, and left my job, I really feel that today.

If you see ads pointing you to votesafe.org to check your voter registration, read this first.

VoteSafe is run by America PAC, Elon Musk's super PAC. It is not a government site. Its own footer says "Paid for by America PAC."

It also doesn't register anyone. The flow ends by telling you to download a PDF, print it, and mail it in yourself.

Its privacy policy (updated Aug 29, 2026) lets America PAC share your info with "business partners" for targeted marketing, or sell it. That includes whether you're registered to vote and "how you submit your vote." Popular Information found your address going to Facebook and Google, a hashed email to The Trade Desk, your usage to X, and a device fingerprint that survives cookie deletion.

VoteSafe also white-labels voter sites for allied groups, including Sentinel Action Fund and Ready to Register. North Carolina's election board warned in July that Ready to Register mailings used an outdated form, went to deceased voters, and had QR codes that could pre-populate other people's data.

To check your registration, use your state's official portal, https://vote.gov, or https://vote.org. None of them sell your data.

Sources:
https://popular.info/p/the-hidden-dangers-of-musks-votesafeorg
https://votesafe.org/privacy-policy
https://www.fec.gov/data/committee/C00879510/
https://www.ncsbe.gov/news/press-releases/2026/07/01/bulk-voter-registration-mailings-cause-voter-confusion-additional-work-counties

Thinking that an LLM can evolve into AGI is like thinking a lawnmower can evolve into a sheep.

In order to boost productivity in the office, everyone walking in the front door will now be assailed by bats. Each bat is trained to encourage usage of a different office feature. “Coffee machine!” “Secure cycle storage!” “Disabled toilet upstairs!” Simply wave your hands around to dismiss the bats and access the workplace proper. You cannot opt out of the bats and they will assail you every time you enter.

This is a post about using any productivity tool in 2026.

If, like me, you are disgusted to discover #1password is funnelling money to fascist garbage fire DHH, the form to let them know your feelings is here: https://1password.com/contact-support

In design I am constantly trying to persuade more junior designers not to move a fence until they understand why it was put there originally. But whoever was in that Cobra meeting clearly looked at the tests for sending an alert and decided the fence (ie the three criteria safeguarding against the cry-wolf effect) was just in the way of a quick and easy action. That is not good governance, that is idiotic misuse.

And the wildfire alerts compounded this by sending *the same alert* multiple times. I know someone who received it five times. In the centre of London. You couldn’t design a scenario more likely to make someone seek out the way to disable them.
And then think of the impact on people who are neurodivergent and may have an increased startle effect, anxiety, misophonia… sending an unnecessary alarm multiple times at random intervals does these people real actual harm.

The reason this matters is the false alert effect. If you send a user alerts that aren’t perceived as important, they do one of two things:
- If they are able, they turn those alerts off (I know people who have done this as a result of the wildfire alerts)
- They ignore future alerts when they arrive
This effect has caused deaths, for example in the clinical setting when prescribers are so used to dismissing routine compliance-driven alerts that they don’t notice a critical dosage warning.

The designers of the UK emergency alert system say the hamfisted wildfire alerts were a misuse of the system https://www.bbc.co.uk/news/articles/ckgw7xj698lo