It’s alive!!!
I have created a GPT!!! Who knew that was a thing? To create one, you have to be in the paying tier – I am, at the lowest level, which is like £7 a month – but you can use somebody else’s if you’re in the free tier.
It’s regular old ChatGPT that you feed your special documents to. In my case, it’s a local history project and I uploaded documents I know haven’t been republished (and therefore AI wouldn’t have them). I won’t point to it, because it’s a little too close to my actual location.
Then you give it instructions for how it is to answer. Tell it what its purpose is. What its preferred online sources should be (you can also switch off its ability to surf the internet). What tone to take. Sample questions. It can have a name and an avatar, y’all.
I could feed it this whole blog and then let it answer questions in my voice. But I won’t do that because there are some seriously bad posts on my blog.
It can be entirely private, open to people who have the link, or open to everybody. The page for exploring other ones is here.
This would be a fantastic idea for someone writing a novel trying to keep character sheets straight. Family history. A complex non-fiction with original research. It’s a blast to play with.
June 18, 2026 — 5:16 pm
Comments: none
Also recommended
I finished the book. I found the early parts more interesting. He concluded with chapters on people who get sucked into AI in an unhealthy way. I came away with the idea that long, continuous conversations (or, even worse, custom-tailored models) are the real danger there. They learn enough about you to get under your skin.
Grok agrees with me.
Anyway, I asked it to recommend a diary program and it recommended a number. I chose Obsidian. It took some fiddling to get it set up to my satisfaction, but now I hit CTRL-N and it opens a new file with today’s date as the filename, in a folder called \2026. I was even able to set up more interesting formatting stuff like make the first line indented using CSS.
It saves them as .md files. Markdown is a simple, human-readable plain text format that uses symbols (like # for headings, ** for bold and * for italics) to add formatting, but it’s easy to read and edit in any plain text editor.
I’ve been meaning to keep a diary for a long time, but it’s becoming imperative as I keep starting projects and losing track of them. I’m kicking myself I didn’t do it for covid. And again for not tracking weather on this soggy island.
But chatting with Grok getting it all set up and it said, “you know if you put your other notes in there, it’ll act like a personal wiki.” No, I did not know that!
I’ll let you know how I get on.
June 17, 2026 — 6:29 pm
Comments: 2
It was a little meta for me
A 2026 book called “How to Talk to AI” came across my threshold today for 99p and I thought, “sure, why not?”
It’s good, actually. As I read, I discussed what he was saying with Grok. It was a very surreal afternoon.
I’m only on chapter two, but interesting stuff. People think LLMs are great at data but terrible at creativity. It’s the opposite – they’re bad on data but fantastic on creativity (if we define creativity as combining different ideas in novel ways). Lateral ‘thinking’ is programmed right in. (Grok agreed).
Give a model two completely different objects and ask how they’re related. It invariably comes up with some interesting and plausible answers that just wouldn’t occur to a human being.
He said that current models are all weighted heavily to the middle. Ask a model to pick a random number between 1 and 10 and it’ll pick 7. I asked Grok. It picked 7. Which means its solutions are corporate and lifeless, particularly when you hear them in large doses.
To get around this, you give the model a persona and ask it to answer as that person. It doesn’t matter who. The examples he gave were “A mad scientist with a penchant for flower arranging. A teenager who is allergic to phones. Elon Musk after smoking weed. A mute nineteenth-century witchdoctor.” It doesn’t change the accuracy of the answer. The act of giving it a role alone forces it to step outside its constraints.
Though they did get even better results on a design challenge when they asked it to answer as Steve Jobs.
Anyway, chapter 2. I’ll let you know how Grok and I get on.
June 16, 2026 — 7:01 pm
Comments: 12
Mucho recommendo
I usually hate ventriloquism acts. Most of them aren’t very good at it, and their dummies are ugly as sin. Most people are creeped out by ventriloquist dummies and they don’t need a Twilight Zone episode to do it.
Then I found Nina Conti. She’s got an emormous range of voices and a flawless delivery.
She doesn’t generally work with dummies, except Monkey. The stuff with Monkey is pretty good. Monkey is dark.
But my favorite is when she gets audience members on stage, straps a mechanical mask to their lower faces, and puts words in their mouths. Some of them are really very funny.
What’s stunning is, these are entirely improvised. She has set routines, but mostly she asks participants what they do and then riff off of it. It’s extraordinary.
Her story goes that she’s an actress and one day on set, her producer gave her a Beginner’s Guide to Ventriloquism as a joke. She threw it in the closet for a year. Later, she ran across it, tried it out and discovered she had a knack for it. That’s how I remember the interview, anyway.
She has a movie out that she wrote, stars and and directed. It’s a road trip, and she’s dressed as monkey. It’s called Sunlight (trailer). I plan to rent it this weekend, so don’t spoil it for me.
June 15, 2026 — 6:49 pm
Comments: 2
Not quite, but interesting
In the newsletter, Bari Weiss says “I Gave a Robot Access to My Apartment.” The actual title of the article is “Training Robots In My Apartment.” Which is closer to the thing. I don’t want to subscribe, so I didn’t read the whole thing, but the teaser was interesting enough.
She allowed a company to send two college students and a chef, to clean her apartment and cook her a meal. This they did for free, wearing cameras. The point of the exercise was to use the resulting film to train robots to do those jobs.
I still think we’re a long way out from functional robots for household tasks. And one of the college kids says to her, “What could happen, in 50 years, is basically all services and goods are free, or are close to free” and I don’t see a way in hell that works out. Who pays for the robot? Who does the maintenace? Still, worth reading the bit you get for free.
Have a good weekend, everyone!
June 12, 2026 — 6:24 pm
Comments: 7
Covered in weasels!
I don’t know the backstory, but this woman seems to be raising six tiny weasels (I had to freeze frame to count them). I bet her meat bill is horrendous.
I was sent the link to that YouTube by @grnspd on Twitter, who I assume is one of you lot.
June 11, 2026 — 4:26 pm
Comments: 7
Don’t piss off the Irish
Photo from this article. This is why they’ve been jailing people for mean tweets. When the dam breaks, it’s going to be epic.
I think. I’ve been wrong before. Sadly, numerous times. I thought the fuel protests of 2000 were the start of something big. I thought the coordinated bombings in London on the 7th of July, 2005 would do it (there was some awful cellphone imagery early on, before it was quashed). I thought no society could absorb the savagery of the Lee Rigby murder. Surely, surely the Southport stabbings would be enough (that’s the one Lucy Connolly went to jail for tweeting about).
Wikipedia has lightly sanitized all those stories (by not posting Rudakubana’s mugshot, for example).
But I think this might be different. Why? The Irish, man.
June 10, 2026 — 5:23 pm
Comments: 5
The Internet Archive just sent me this
It’s the photo that started the series of online creepypastas in 2019 that turned into the recent movie Backrooms in 2026 that’s doing all kinds of excellent at the box office. According to the email, the photo was actually of a 2003 renovation of a former furniture store in Oshkosh, Wisconsin.
I do hope this kind of thing is the future of film. Stupid Hollwood – refuses to make films people want to see.
I think I was peripherally aware of backrooms at the time, being a terminally online kindofa mustelid. The image is supposed to represent a creepy and endless interdimensional space. You wander through and nothing ever happens, except it sometimes does. Or corner-of-the-eye hallucinations.
I feel like it was a simulation at one point, but I may be confusing it with the Stanley Parable. But the Stanley Parable came out eight years before, so no. Similar feel, though.
June 9, 2026 — 5:57 pm
Comments: 4
More AI toys
I recorded a meeting on my phone between three people. I needed a transcript. There are several online services, but I tried a couple and wasn’t at all happy.
The the robot suggested I download an app called Vibe, powered by OpenAI’s Whisper. It lives locally on your computer. You feed the audio into it and a transcript comes out the other end (after a time, naturally).
It did a fantastic job. It didn’t label the three participants, which was a pity, but the text was clean and nearly perfect (a few proper names off and some rather bizarre repetition of one-liners).
Uncle B is off sobbing in the corner. So much of his career was spent doing interviews – and then, o god, transcribing them by hand.
June 8, 2026 — 7:17 pm
Comments: 3
Chickens for scale
I told you this rose was about to explode (here it is in color.
Positively the last chicken post of the week. Have a good weekend!
June 5, 2026 — 5:55 pm
Comments: 7