Daniel sent us this prompt about the history of children's TV — and it's one of those questions that seems simple until you actually sit with it. Have we ever had a consistent principle for what's appropriate for kids? How do shows like Teletubbies, the Morbegs, and Rosie and Jim stack up against the broader repertoire? And then the kicker: what happens when AI starts generating content for the under-seven crowd on platforms like YouTube Kids? It's a question about nostalgia that turns into a question about the future of childhood itself.
The timing is right for this. We're at this inflection point where the economics of children's content have completely inverted. A human-produced episode of Bluey costs somewhere around two hundred thousand dollars per seven minutes. An AI-generated equivalent using Sora two point oh or Pika three point oh costs about twelve dollars in compute. That's not a gap, that's a cliff.
So the question isn't whether AI-generated children's content will flood the market. It already is. The question is what gets lost when the cost of production drops to zero and the only signal that matters is watch time.
Before we can understand what AI is doing to children's TV, we need to understand what children's TV was supposed to be in the first place. And the history here is messier than most people assume.
I think a lot of people assume children's television has always been this safe, educational, carefully vetted space. And that's just not true. Early children's programming in the United States — Howdy Doody, which launched in nineteen forty-seven — was essentially a toy commercial. It was built to sell products to kids. The whole educational mandate only really emerged in the nineteen sixties with Sesame Street and the Children's Television Workshop.
Even that mandate was a reaction to a specific cultural panic. People forget that Sesame Street was created in nineteen sixty-nine partly in response to the finding that children from lower-income households were entering kindergarten with dramatically smaller vocabularies than their middle-class peers. The show was designed as a targeted intervention — a televised head start program. It wasn't just "let's make something nice for kids." It was "we have data showing a systemic failure, and television might be the most efficient way to reach millions of children simultaneously.
And Joan Ganz Cooney, who founded the Children's Television Workshop, was not a television producer. She was a documentary producer and journalist who had spent years studying educational inequality. She approached children's television as a public health problem. Every segment of Sesame Street was tested for educational outcomes before it aired. They had a research department that was as large as the production department. That model — research-driven, publicly funded, pedagogically intentional — set the template for what we now think of as quality children's programming.
The hidden curriculum of children's media, and who gets to write it. That's what this is actually about.
And if you look at the history, you can roughly divide it into three eras. First, the Public Service Era, roughly the sixties through the nineties. This is when governments and broadcasters treated children's programming as a public good. The BBC had a whole children's department with developmental psychologists on staff. The second era is the Commercial Cable Era, nineties through the twenty-tens — Nickelodeon, Disney Channel, Cartoon Network. Still human-produced, still had creative gatekeepers, but the business model was advertising and merchandising. And then the third era, which we're in now, is the Algorithmic Feeds Era. YouTube Kids launched in twenty fifteen, and suddenly the gatekeeper isn't a commissioning editor with a degree in child psychology — it's a recommendation algorithm optimizing for retention.
Three eras, three different theories about what a child actually is. And that's where the three shows in the prompt come in. Teletubbies, the Morbegs, and Rosie and Jim. Each one represents a completely different answer to the question of what children need from television.
Let's look at three shows that each had a very different answer to that question. And I mean very different.
Start with Teletubbies. This is a show that a lot of people my age remember as this bizarre, almost psychedelic experience — the sun-baby, the tummy screens, the repetitive language. But what most people don't know is that every single element of that show was designed using actual research on infant gaze patterns and language acquisition.
Anne Wood and Andrew Davenport, the creators, spent years developing the concept. Wood had been a secondary school English teacher, and Davenport had studied speech science and had a background in early childhood development. They produced a production bible that was a hundred and twenty-seven pages long. Every character's movement pattern, every color choice, every sound — all intentional. The repetition that adults found maddening? That was the point. Pre-linguistic children learn through repetition. The slow pacing? Designed to match the processing speed of a twelve-to-eighteen-month-old brain.
They didn't just guess at that pacing. They actually brought toddlers into a viewing lab and tracked their eye movements. If children looked away at a particular moment, that segment was re-edited. The show was essentially co-designed by its audience, just without the audience knowing it. The show ran from nineteen ninety-seven to two thousand one in its original run, three hundred and sixty-five episodes, total budget about two and a half million pounds. They tested the pilot on two hundred children in Birmingham over six months before they aired a single episode.
The controversy in nineteen ninety-nine over Tinky Winky being quote unquote gendered — Jerry Falwell famously called the character a gay role model — that whole moral panic was a complete distraction from what was actually innovative about the show. Teletubbies was television as a mirror for pre-linguistic cognition. It wasn't trying to teach facts. It was trying to teach pattern recognition, emotional regulation, the joy of repetition itself. The tummy screens were a deliberate acknowledgment that even toddlers in nineteen ninety-seven were growing up in a media-saturated environment. The show was saying, we know you see screens everywhere, so we're going to put a screen inside the screen and make that part of the learning.
The glockenspiel of corporate approachability, but for babies.
That's actually not far off. And here's the key thing: Teletubbies had a theory of the child baked into its DNA. The theory was that very young children are pattern-recognition machines. They need repetition, they need slow pacing, they need visual and auditory cues that help them map language onto experience. Whether that theory is correct is debatable — but it was a theory, and it was tested.
Now contrast that with the Morbegs. This is a show that most of our listeners outside Ireland will have never heard of, which is a shame because it might be the most psychologically sophisticated children's program ever produced.
The Morbegs was produced by RTÉ, the Irish national broadcaster, with funding from the Irish Department of Education. It ran for three series from nineteen ninety-five to nineteen ninety-eight, fifty-two episodes total. And it was created in direct response to a nineteen ninety-three government report that found Irish children had higher rates of anxiety than their European Union peers.
The government literally commissioned a children's show as a mental health intervention.
And the show they got was deeply surreal. The Morbegs were these two large, colorful, somewhat shapeless creatures — the name comes from the Irish word "mór" meaning big, and "beag" meaning small. They lived in a castle in the woods, and they were pursued by this terrifying entity called the Burning Tree. The show used elements of Irish folklore and deliberately unsettling imagery to help children process fear in a controlled environment.
I remember watching this as a child and being genuinely unsettled. But I also couldn't stop watching. And I think that's worth unpacking for a moment, because there's a real neurological basis for why this approach might work. When a child encounters something frightening in a narrative context — in a story, on a screen, in a book — their amygdala activates, the fight-or-flight response kicks in, but the prefrontal cortex stays online because the child knows, on some level, that they're safe. They're sitting on a couch, they're in their living room, there's a parent nearby. That gap between the amygdala's alarm and the prefrontal cortex's reassurance — that's where emotional regulation skills get built.
That was exactly the intended response. The creators — Irish language activists and child psychologists working together — believed that children need catharsis. They need to encounter fear in a safe container where they can practice managing it. The faceless Morbegs, the ominous forest, the sense that something threatening was always just off-screen — this wasn't poor production design. This was trauma-informed storytelling, decades before that term entered the mainstream.
It connects to a much older tradition of children's stories that we've sort of sanitized out of modern culture. The original Grimm's fairy tales are horrifying. Children get eaten. Children get abandoned in forests. And those stories survived for centuries because they served a psychological function. The Morbegs was doing something similar, but for the television age, and with actual psychological research backing it up.
It was never exported. The show was considered too dark for international markets.
Which is fascinating, because what does "too dark" mean here? It means adults in other countries looked at it and felt uncomfortable. The children weren't consulted. The adults made the gatekeeping decision based on their own discomfort.
Which tells you something about how different cultures define "appropriate." In Ireland in the mid-nineties, after a government study showed elevated childhood anxiety, the appropriate response was to create a show that helped children sit with discomfort. In other markets, the appropriate response was to avoid discomfort entirely. Same children, same developmental needs, completely different cultural judgments about what serves those needs.
Then you have Rosie and Jim, which is almost the anti-Morbegs. No fantasy, no surrealism, no psychological experimentation. Just two ragdolls on a narrowboat, learning about canals and locks and wildlife.
Rosie and Jim ran for ten series from nineteen ninety to two thousand, a hundred and thirty episodes, all filmed on an actual narrowboat on the UK canal system. John Cunliffe created it through Ragdoll Productions — the same company behind Teletubbies, interestingly — and the concept was beautifully simple. Two ragdolls come to life when nobody's looking, and they explore the world around them with the boat owner, who never notices they're alive.
There's something almost radical about that simplicity when you compare it to what children's television became later. No rapid cuts, no manic pacing, no overstimulation. Just a narrowboat moving at four miles an hour through the English countryside, with a gentle voiceover explaining what a lock is and how it works. The creative constraint was the boat. The boat forced discipline. Every episode had to take place on or near the water, which meant the show couldn't sprawl into fantasy or abstraction. Children learned about concrete things: how locks work, what ducks eat, why some bridges open and others don't.
That constraint produced remarkable consistency. A hundred and thirty episodes over ten years, and the quality barely wavered. Rosie and Jim had a theory of the child too: children need concrete reference points. They need to see real things in the real world and hear them explained by a calm, friendly adult. No moral lessons, no emotional processing — just gentle, factual engagement with the physical environment.
You've got three shows, three theories. Teletubbies says children are pattern-recognition machines who need repetition and sensory calibration. The Morbegs says children are emotional beings who need controlled exposure to fear and uncertainty. Rosie and Jim says children are empirical learners who need concrete facts delivered in a calm, predictable container. And what strikes me is that all three of these theories are defensible. You can make a case for any of them, depending on the child and the context.
None of these theories is provably correct. They're creative hypotheses, tested against audience data that took years to collect. The creators of all three shows would probably admit they were making educated guesses. But they were guesses made by people who had spent their careers thinking about what children need, and who had the time and resources to test their assumptions.
Which brings us to the present. Because AI-generated children's content has no theory of the child at all. It has a theory of engagement metrics.
Those three shows represent the old model: expensive, intentional, and slow. Now let's talk about what happens when you remove all three constraints.
Let's start with the scale of the problem. YouTube Kids, since its launch in twenty fifteen, has become the primary content platform for a generation of children. By twenty twenty-three, it had over thirty million active monthly users in the United States alone. And the platform has hosted over thirty million children's videos since twenty twenty.
Thirty million videos. To put that in perspective, the entire original run of Teletubbies was three hundred and sixty-five episodes. The entire run of Rosie and Jim was a hundred and thirty. The Morbegs was fifty-two. Humanity's entire canon of carefully produced children's television could probably fit into a single day's uploads on YouTube Kids.
The quality control is essentially non-existent. A twenty twenty-three study by the University of Michigan found that forty-two percent of content labeled as educational on YouTube Kids contained no educational value whatsoever. A twenty twenty-four study by the Oxford Internet Institute found that sixty-seven percent of children's videos on the platform contained at least one instance of what they called disturbing or age-inappropriate content.
"disturbing" here doesn't necessarily mean violent or explicit. It means content that is psychologically unsettling in ways that young children can't process. Rapid cuts, distorted faces, screaming, characters in peril with no resolution — the kind of sensory assault that hijacks a toddler's attention without providing any narrative or emotional structure.
This is the Elsagate problem, but evolved. For listeners who don't remember, Elsagate was the twenty seventeen scandal where YouTube was flooded with videos featuring popular children's characters — Elsa from Frozen, Spider-Man, Peppa Pig — in bizarre, often violent scenarios. Characters eating each other, characters being buried alive, characters undergoing medical procedures. The content was algorithmically generated or hastily assembled in overseas content farms, and it was racking up millions of views because children would watch it to completion, transfixed by the sensory overload.
YouTube responded with stricter content moderation after the media firestorm. But the moderation is keyword-based, not concept-based. You can't use certain words in titles or descriptions, but you can generate anything you want visually. And with AI video generation tools — Sora two point oh from OpenAI, released in December twenty twenty-five, can generate four-K video up to sixty seconds from a text prompt at about fifty cents per minute of output — the barriers to creating this content have essentially vanished.
Fifty cents per minute.
A twenty-minute video for ten dollars. A content farm can generate hundreds of videos a day, each one slightly different, each one A-B tested by the algorithm against millions of toddlers, each one optimized for the one metric that matters: did the child keep watching?
This is where the tragedy of the commons kicks in. When production costs drop to zero, the only signal that survives is engagement. Not educational value, not developmental appropriateness, not emotional safety — just retention. Did the child look away? Did the child click on something else?
There was a case in twenty twenty-four where an AI-generated Blippi clone channel — and Blippi, for those who don't know, is this incredibly popular children's entertainer who does educational videos — this clone channel accumulated eight hundred million views before it was taken down. Eight hundred million views. The algorithm promoted it because children watched it to completion, and children watched it to completion because it was designed to be unputdownable, not because it was good.
The musical equivalent of beige wallpaper, but optimized by a superintelligence.
Here's the thing that keeps me up at night. A twenty twenty-two study showed that parents correctly identified inappropriate children's content only fifty-four percent of the time. Barely better than a coin flip. So the idea that parents can just screen what their kids watch — it doesn't hold up. The content looks like children's programming. It uses the same characters, the same color palettes, the same musical cues. The difference is in the structure, the pacing, the emotional arc — things that are hard to notice unless you're watching carefully, which most parents aren't doing for hours at a time.
Even if a parent is watching carefully, the content is designed to exploit the gap between adult perception and child attention. An adult sees a brightly colored train and hears a cheerful song and thinks, "This is fine." But what the child is experiencing is a carefully engineered attention trap — no narrative resolution, no emotional calibration, just a continuous dopamine drip of novel stimuli. The parent's brain processes the surface; the child's brain is being shaped by the structure.
The gatekeeping function has moved from commissioning editors to algorithms, and neither parents nor regulators have the tools to evaluate what's coming through.
The regulatory framework is completely mismatched to the problem. COPPA, the Children's Online Privacy Protection Act from nineteen ninety-eight, regulates data collection from children, not content quality. The UK's Age Appropriate Design Code, which came into effect in September twenty twenty-one, is about data protection and privacy by design. The FCC's Children's Television Act of nineteen ninety mandated educational programming quotas, but those rules apply to broadcast television, not streaming. There is no regulatory framework anywhere in the world that evaluates whether a children's show is developmentally appropriate.
We're in this bizarre situation where a children's app can't collect your kid's email address without consent, but it can show them an AI-generated video of a cartoon character screaming for eight minutes, and that's perfectly legal.
Perfectly legal, and in fact algorithmically rewarded. Because the screaming video probably has better retention metrics than the calm, educational one.
Let me ask you something. The prompt asks what kind of creatives are drawn to making children's content. Looking at the history, what patterns do you see?
The BBC Children's archive has interview data going back decades, and it's fascinating. You see three archetypes showing up again and again. First, the pedagogue — the person who wants to teach. They see children's TV as a classroom, a way to transmit knowledge and skills. Andrew Davenport from Teletubbies is a pedagogue. Second, the nurturer — the person who wants to protect. They see children's TV as a safe space, a refuge from a scary world. John Cunliffe from Rosie and Jim is a nurturer. And third, the anarchist — the person who wants to subvert adult expectations about what children should experience. They see children's TV as a space for creative risk-taking that the adult media landscape no longer allows.
The Morbegs creators were anarchists.
They looked at a government report about childhood anxiety and said, the solution is to scare children more, but in a controlled way. That's an anarchist move. It's also probably the right move, given what we now know about how children develop emotional resilience. But it's not a move that a committee would make, and it's certainly not a move that an engagement-optimizing algorithm would make.
AI has no archetype. It has a prompt.
When you prompt an AI to make a children's video, you get output that looks like children's content. But there's no pedagogue making decisions about what to teach and in what order. There's no nurturer thinking about whether this image might be frightening. There's no anarchist pushing boundaries in ways that serve the child rather than the algorithm. There's just the prompt and the training data and the optimization target.
Compare the production bible for Teletubbies — a hundred and twenty-seven pages, every character's movement pattern specified, every color calibrated — with a typical AI prompt: "make a kids video about a train that is happy." That's it. That's the entire creative document.
The AI will make that video. It'll be bright, it'll be colorful, the train will smile, there'll be cheerful music. A toddler will watch it. But there's no theory underneath it. No hypothesis about what the child needs. No testing against developmental outcomes. It's content that looks like children's television the way a plastic apple looks like food — all surface, no nutrition.
Let me push on that analogy, because I think it's actually more precise than it first appears. A plastic apple isn't just missing nutrients. If a child tries to eat it, it can cause harm. The AI-generated children's video isn't just empty — it's occupying time and attention that could have been spent on something developmentally nourishing. Every minute a toddler spends watching an AI-generated attention trap is a minute they're not spending watching something with narrative structure, or emotional calibration, or factual content. The opportunity cost is the harm.
Where does this leave us? This sounds like a disaster, and in many ways it is. But there are things we can actually do about it.
Let me offer three concrete insights, aimed at three different groups. For parents and educators: the single most reliable signal for quality children's content is production transparency. Shows that publish their educational rationale — like Teletubbies' research papers or the Morbegs' government reports — are almost always better than shows that don't. If you can't find any information about who made a show and why they made it, that's a red flag. Demand transparency from platforms. Ask YouTube Kids to label content that has been reviewed by human child development specialists.
For creators — people who are actually making children's content with AI tools. What should they be doing differently?
Embed a theory of the child into your prompt engineering. Don't just specify the visual style and the topic. Specify the developmental stage you're targeting — is this for a twelve-month-old who needs sensory calibration, or a four-year-old who needs narrative structure? Specify the attention span you're designing for — are you making a two-minute video for a child who can focus for ninety seconds, or a ten-minute video for a child who can follow a simple plot? Specify the emotional arc — does this video start calm, introduce mild tension, and resolve it? Or is it just stimulation from start to finish? The best AI children's content will be the most constrained, not the most generative.
Constraint as a feature, not a bug. Like the narrowboat on Rosie and Jim.
The narrowboat forced discipline. AI tools need the equivalent of a narrowboat — a creative constraint that prevents sprawl and forces coherence.
The twenty twenty-six reauthorization of the Children's Television Act should extend educational programming requirements to streaming platforms. This is the year to do it. And it should include a developmental appropriateness standard that can be audited. We have the technology to evaluate content at scale — we can train classifiers to detect rapid pacing, disturbing imagery, lack of narrative structure. This is a political will problem, not a technical one.
The FCC's nineteen ninety rules were written for a world of broadcast licenses and Saturday morning cartoons. They're completely irrelevant to a world where a twelve-dollar AI-generated video can reach eight hundred million views. And I want to underscore how fundamentally the economics have shifted. Under the old model, if you wanted to make children's content, you needed funding. You needed a broadcaster to commission you, or a studio to back you. That created a natural filter — not a perfect one, but a filter nonetheless. Someone with resources and reputation had to sign off. Now, the filter is gone. The only remaining filter is the algorithm, and the algorithm optimizes for one thing.
Here's the thing about the Morbegs that I think is the most important lesson in this whole conversation. Sometimes the best children's content makes adults uncomfortable. The Morbegs was too dark for international markets. Adults found it disturbing. But children loved it, and it was designed specifically to address a documented mental health need. If an AI system is trained on adult preferences for what children should watch — and most content moderation systems effectively encode adult preferences — it will produce safe, bland, pedagogically empty content. We need to preserve the space for creative risk-taking in children's media, even when it makes us uncomfortable.
The anarchist archetype. You can't automate the anarchist.
You can't. And you probably shouldn't try. The anarchist needs to be a human who understands children well enough to know which boundaries are worth pushing. An AI doesn't understand children. It understands patterns in training data. And the training data for children's content is overwhelmingly the safe, the sanitized, the commercially successful. The Morbegs wouldn't survive an AI training process. It would be an outlier, a statistical anomaly, something to be smoothed away.
Where does that leave us? Let me leave you with one final thought about the sun-baby.
The sun-baby from Teletubbies — the actual infant whose face appeared in that sun, giggling down at Tinky Winky and Dipsy and Laa-Laa and Po — that baby is now twenty-nine years old. She's been walking around in the world for nearly three decades. What did that show teach her? Pattern recognition, probably. Emotional regulation, maybe. The joy of repetition, almost certainly. The show had a theory about what she needed, and it spent years testing that theory, and it produced three hundred and sixty-five episodes that were coherent with that theory.
Today's toddler is watching an AI-generated Peppa Pig drive a car into a swimming pool and scream for eight minutes. What is that teaching her? We don't know. And the fact that we don't know — that there's no theory, no testing, no intentionality behind it — that's the problem.
As AI-generated content becomes indistinguishable from human-made, will we develop new heuristics for trust? Will we learn to look for the signs of intentional design the way we learn to look for the signs of a well-built chair? Or will we simply accept that children's media has become a purely algorithmic experience, like search results or social feeds?
I think that's the open question. And I don't think we have an answer yet. What I do know is that the history of children's TV is a history of adults trying to figure out what children need. It's a history of pedagogues and nurturers and anarchists making bets about the developing mind and testing those bets over years. AI doesn't need to figure anything out. It just needs to keep them watching. That's not a technology problem. That's a values problem.
Values are the one thing you can't prompt-engineer.
Values are the thing you have to bring with you. And right now, we're not bringing them to the most important media environment in the world — the one where our youngest children are spending their most formative hours.
Now: Hilbert's daily fun fact.
Hilbert: In the late Victorian period, a British expedition to Madagascar discovered a lava tube ecosystem where blind cave-dwelling crickets had evolved to feed exclusively on the roots of fig trees that penetrated the ceiling from above, creating an entirely self-contained food web that had never seen sunlight.
...right.
This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop for the fact that will now live in my head rent-free. If you enjoyed this episode, please rate and review the show — it helps other curious minds find us. I'm Corn.
I'm Herman Poppleberry. We'll be back next time.
