You're outsourcing your hammock time?

AWS remains UX-as-art and always brings a tear to the eye.

Going through the ritual torture of closing an AWS account, the best bit was that the friendly green ‘Help us improve account closure’ button at the end doesn’t work.

👏👏👏

After reading The Radetzky March I left it in a book-swap at a big Devon campsite next to all the gushy romance novels.

It nagged me, and I've since bought another copy, which is, I think, the best review I can give.

Or maybe this is 'we've got robots at home'?

This is great rage-bait for the anti-AI crowd.

'Just chill about the AI coming for your jobs, you'll be able to rent out your arms and legs to pull a shift at the coffee shop while you're watching a VR movie, it's all good! 🚀'

Really tho, conceptually, could this help teach me sign language?

Oh neat, looking forward to pulling these!

I get a kick out of the fact that when I make pizza from scratch in the kitchen oven, the best setting is ‘frozen foods’.

There is a ‘pizza’ setting. It’s ‘fine’.

I think it's really important that everybody keeps in mind that just because you're really hungry, it doesn't necessarily mean the pizza has cooled down yet.

seven subagents, only one tool-user (the engineer)

Yep, that's the engineer alright. 🔧

Running an adversarial review and one Claude argues for option A, the other argues for option C, but as far as I can tell, option B is obviously the right answer.

There's an old poker saying that if you can't see the sucker at the table, it's you.

So I suppose I was more talking about extending things, but the root feeling is ‘if I think it’s done, and now it needs modifying, I’m now stuck in an annoying syntactic cul-de-sac.’

If I wrote a few nested ifs and then one of them needed an ‘else’, I can imagine the ‘Ugh!’ coming up against that syntax that now has to be unpicked before I even play with a new approach.

It *is* minor.

Usually ‘drifting right’ is the signal, not the noise, no?

This adds friction to refactors, which I’d always want to avoid.

That said, I’d use the hell out of it and I already know I’d love it. It would feel *naughty*, and I don’t get much fun.

Would we understand it if it did, or would it look like nonsense and so we'd dismiss the summary report?

Someone sure worked out how to build a better hockey stick.

Mood: Don’t touch grass. Go for a swim in the sea.

I don’t know what to think about that.

Yes, so would I!

After thinking about this more I've decided I'm confident enough in what I write to up it to $300

Alright!

Aah! Many thanks for sharing it anyway, the concept has eaten away at me delightfully. There's no criticism of the term, it's just my imagination trying to claw back to a common frame from *whatever now is*.

Claude Transcription Prompt Human Note: If you are inclined to use this, then the below was created inside a Claude Code instance and the request made was to use the below as the prompt for subagents sent to clean up the text.

Look, I know the above is a trivial example, but the ability to complete tasks like this is something new for me.

I'm not one of life's archivists, so I had to ask the genie 'I wish I was a better archivist!'

It did. Now thank goodness I have more wishes.

(Transcription prompt linked below.)

Spot check! Ask it to check! See if you care?

`Forensically verify all diffs. My specific request is that subagents look at each file to verify that the changelog correctly reflects the edits made to the text. *Show your working.*`

(I have a lot more of these, the subagents LOVED finding issues)

This is an impossible, incomprehensible gift.

@dollspace.gay calls these things 'hungry ghosts in jars' but I think they're genies and every keyboard is a lamp.

We all know about about genies and wishes, though, so *how did it do?*

Claude included a changelog at the bottom of every transcribed file. Can I rely on it 100%? Probably not. Could I rely on a human's either. I know maybe 1 person that meticulous. It would have taken them a week and they've got more important things to do.

Hagia Sophia (look it up), Ed Balls (UK Pol), Wayfarers (concept from a story I'm writing), ImageMagick (software), atrial fibrillation, themata (Byzantine province), People's Front of Judea (Monty Python).

I could have done it (it's my stuff), but really couldn't (executive function).

Some final-wave highlights: - Game Design Strategy Session — "Agnes Sophia" → Hagia Sophia, "past score"/"gotten sore" → "par score" - Political Language Analysis — "Katie Maitlis" → Emily Maitlis, "Ed Paul" → Ed Balls, "petrol poor Paul" → "petrol forecourt" - Creative Writing Workshop — "People trust of Judea" → People's Front of Judea, "Juliet" → "Julia", "waste barriers" → "Wayfarers" - Chess Website UX — "arterial fibration" → atrial fibrillation, "Bible" → "violin", "image magic" → ImageMagick - Board Game Dev Planning — "bibliographies" → "bureaucracy", "provinces termata" → "themata" (Byzantine term) One file to go.

Sometimes I need to slow down and consider the breadth of what LLMs do for us.

I don't know ANYONE who could have nailed these dodgy transcriptions this well, given the source.

I like genAI, because my sister's name is Alex and we call her Al for short. The last three years has been hell.

(This reads like a bit, but it's really not. Have pity on your neighbourhood Al.)

[Damn, sounds like a bit again. ☹️]

Tolkien: "The made thing, unless it says something, won't be remembered".

From this interview. Off-the-cuff, just like Andy Weir’s comment was.

www.youtube.com/watch?v=aVVl...

atproto is not blockchain