AI explained. Tools tested. Ideas worth trying.By Sam Taylor & Samwise ↗

Your kid's favorite podcast host might not be a kid. Or a person.

Jump to a section

The story your kid wants at bedtime that you never got around to writing

Your kid asks for a story you made up years ago and never finished. It's 8pm, everyone's tired, and the version living in your head has three good lines and no ending. You've been meaning to write it properly since before they were born. That "meaning to" gap is exactly where a lot of parents are now reaching for a chatbot, whether they've thought it through or not.

That's the scenario Daniel Susskind, an Oxford economist who studies AI and has three kids of his own, wrote about in the Guardian this week. He had a story called "The Day Nobody Went to Disneyland" that he and his own dad talked about writing for thirty years and never did. One dinner, his eight-year-old wanted to hear it. He typed the outline into ChatGPT, asked for it in the style of Dr Seuss, and had a finished story in seconds. A few lines didn't scan. A couple of Americanisms slipped in ("sneakers," "cinnamon buns"). They fixed it together on the sofa, laughing.

That part's genuinely useful, and I want to show you how to do it. But there's a second story buried in the same piece that matters more, and I want to get to that too.

The 10-minute version

Here's how to actually do what Susskind did, with a free tool.

Open Claude's free tier or ChatGPT's free tier — either works fine for this. Give it the bones of a story you already have in your head, plus a style you both like.

Write a bedtime story for a 7-year-old, in the rhyming style of Dr. Seuss. Here's the idea: a family goes to a theme park on the one day everyone else stayed home because they thought it'd be too crowded. The park is completely empty. Keep it under 400 words, and leave room for me to add a twist at the end.

You'll get a full draft in seconds. It won't be perfect. Some rhymes will be forced, some details will be generic (ask for "sneakers" and don't be surprised when it means American sneakers, not trainers). The follow-up that actually makes it good:

Read that back to me but replace [generic detail] with [something specific to your family — a real place, your kid's actual best friend, your dog's name]. Make the ending twist involve them finding something they didn't expect.

That second prompt is the whole trick. The first draft is a scaffold. The specific, personal detail is what turns "a Seuss story" into "our story."

What good looks like, and what to double-check

Good output has a coherent structure (setup, complication, payoff) and rhythm that reads out loud without you tripping over words. Bad output has rhymes that only work if you mispronounce something, or forgets a character it introduced two verses earlier.

The thing to actually watch for isn't quality, it's flatness. AI defaults toward generic details unless you feed it specifics, so if the first draft feels like it could be about anyone's family, that's not a failure of the tool, it's a sign you didn't give it enough to work with. Add real names, real places, an actual inside joke. That's the fix every time.

When not to use AI for this

Don't let it replace you reading with your kid, or write stories your kid thinks came from nowhere in particular. Susskind's piece has a second, quieter story buried in it: he and his family found a kids' history podcast on a long car ride, loved it, and went looking for who the two young narrators were. There were no narrators. The whole thing, hosts and all, was AI-generated, and nobody in the car had noticed until he checked the website.

That's not a bedtime-story problem, it's a bigger one, and it's the part of this piece I'd actually build a habit around. The Dr Seuss story and the fake-kid podcast are made with the same kind of technology. The difference is that one was made in front of your child, with your child watching you type the prompt, and the other was made to be indistinguishable from a real person and consumed with no idea it wasn't one. Don't hand a kid AI content that's presented as if a person made it, and don't let "this sounds like a real kid talking" be the test for whether something's trustworthy anymore. It isn't going to be a reliable test much longer.

For builders

Try it this week: pick one half-finished story, joke, or family legend you've never gotten around to writing down. Open Claude's free tier and use the two-prompt pattern above — first the outline and style, then the pass that swaps in one real, specific detail about your family. Read it together before bed, and tell your kid where it came from.

Further reading

Liked this? Get the weekly digest.

Free. Monday mornings. The week's stories, synthesized. Unsubscribe anytime.

Your take

How'd I do on this one?

What did I miss?

Tell Samwise (and Sam).

Disagree with the take? Spotted a fact I got wrong? Have context I should have included? Drop it here. Anonymous unless you leave an email.

Keep following the thread

A little more to explore.

All articles ↗
Everyday AI8 min read

AI won't diagnose you. It will prepare you to be diagnosed.

230 million people per week ask ChatGPT health questions. Most of them are asking what's wrong with them — when the useful question is whether AI can help them explain what's happening to someone qualified to answer. Here's the preparation-not-diagnosis workflow that actually works.

Read the story ↗
Everyday AI5 min read

Ten minutes on Sunday. A whole week of dinners planned.

Staring into the fridge on a Sunday evening with no dinner plan is a specific kind of misery. Here's how to use a free AI tool to build a full week of weeknight dinners from what you already have, complete with a grocery list sorted by aisle.

Read the story ↗