# Using Jev as a teacher to help an SLM write better stories > Source: > Published: 2026-09-20 17:02:00+00:00 # Giving a tiny story-writer a teacher For a while now I've been playing with a slightly odd question: **how small can an AI be and still tell a story?** The AI systems everyone talks about — the ones behind chatbots — are enormous. They have tens or hundreds of *billions* of internal settings (people call these "parameters" — think of them as tiny knobs the model tunes as it learns). That size is why they're expensive to run and impossible to fit on a phone or a toy. I wanted to go the other way. How much can you strip away and still get something that writes a real, if wobbly, little story? ## A storyteller the size of a rounding error So I built one. I shrank the vocabulary down to the simple words a small child uses, cut the model's size again and again, and trained it on **TinyStories** — a collection of very simple tales written with a toddler-sized vocabulary, made exactly for this kind of experiment. The result is a storyteller with about **14 million** parameters. That sounds like a lot until you remember the big models have *a hundred thousand times* more. And it works — sort of. It writes recognizable little stories about girls and boys and balls and parks. They're charming. They're also, frankly, not very good. It loses the thread, repeats itself, introduces a character and forgets them, has things happen for no reason. Here's a real example of what it produced on its own: …they also liked to pretend they were explorers and pretend they were explorers. …ben had a sad face and a hole in its fur… A stutter ("pretend they were explorers" twice), and a friend who somehow has fur. Fun to read, but not a story you'd want read to you. That's roughly where you'd expect a model this tiny to top out. But I didn't want to stop there. ## The insight: a child doesn't write alone Here's the thing that nagged at me. A four-year-old tells stories with a *tinier* vocabulary than my model — fewer than two thousand words — and their stories are often better. How? Part of the answer is that **a child rarely writes alone.** There's usually a parent or a teacher nearby, gently nudging: *"Wait — why did the horse start singing?"* … *"What did the forest look like?"* … *"You said the leaves were soft *and* sharp — which is it?"* The child doesn't need those questions written into their head. They just need someone to ask them at the right moment, and they correct themselves. So I wondered: **what if my tiny writer had a teacher too?** ## Enter the teacher The teacher I reached for is a model called **Jev**. Jev is unusual: it doesn't write anything. What it does is *judge*. You show it a piece of text and ask it precise questions — "Does this sentence follow from what came before?", "Is this too repetitive?", "Did this event happen for a reason the story gives?" — and it answers each one with a **calibrated confidence**, a number saying how sure it is. It's less like a writer and more like a very fast, very consistent reading teacher who only ever answers yes/no questions, honestly. That turns out to be exactly the shape of a teacher you can put in a loop. So here is what my tiny writer now does. It writes **one sentence**. Jev reads the whole story so far and checks it against a handful of those questions. If everything's fine, the writer moves on. If Jev flags something — the sentence drifts off topic, or repeats a word, or an event came out of nowhere — the writer simply **tries that sentence again**, and again, until it passes. Then it writes the next one, with the teacher watching the whole time. ## It actually works Here's a single correction, caught live. The writer produced: ✗ *she wanted a band of different colors.* Jev flagged it — not for grammar (it's fine) and not for being off-topic, but for being **unmotivated**: a vague wish that leads nowhere. ("A band of different colors" — you can't quite picture it, and nothing follows from it.) So the writer tried again: ✓ *she had a big, red ball that she loved to play with.* A concrete thing, one the rest of the story can actually use. The teacher's own coherence rating for the story jumped from a 2 to a 3 out of 3. Multiply that across every sentence and the difference is night and day. Remember the stuttering, fur-covered mess from before? With the teacher in the loop, the same little model wrote this: once there was a little girl named lily. she had a big, red ball that she loved to play with. one day, she went to the park to play with her ball. she was very happy and ran around all day. but then, she tripped on a rock and hurt her knee. "oh no, lily!" she cried. That's a *real little arc* — a setup, a mishap, a reaction — instead of a stutter and a friend with fur. Same tiny brain. The only thing that changed is that someone was asking it the right questions as it went. ## Why I think this is a big deal Sit with what just happened: a model a hundred-thousandth the size of a frontier system produced a genuinely better story — not by getting bigger, but by getting **coached.** The intelligence didn't all have to live *inside* the writer. Some of it lived in the teacher standing next to it. That opens up two directions I find genuinely exciting. The first is about **doing more with less.** If a small, cheap model plus a good teacher can punch this far above its weight, you don't always need the giant expensive model. You need a decent little one and a sharp critic — and a critic that only answers focused questions can be small and cheap too. (I want to be honest about the ceiling: the teacher can only ever pick from sentences the little writer is *capable* of producing — it can't put words in its mouth it could never have found. It makes the writer reliably do its best, not become something else. But "reliably its best" is a large improvement.) The second is about **people, not models.** Look again at that loop — a learner produces something, a patient teacher reads it and asks exactly the right question at exactly the right moment, and the learner revises. That's not just how you coach a tiny AI. **That's what good teaching is.** The same idea — a tireless, consistent tutor that reads a student's own writing and nudges with the right question instead of just marking it wrong — could sit inside education products and help real children write better. That's the version of this I most want to explore. ## What's next: learning, not just redoing There's one honest limitation in what I've described. Right now, when the teacher flags a sentence, the writer just **rewrites it** — in that moment. It doesn't *remember* the lesson. Ask it to write a fresh story tomorrow with no teacher, and it'll make the same kinds of mistakes again. Real learning is the opposite: you internalize the feedback so you stop needing it. I've taken a first step in that direction — I fed the teacher-corrected stories *back into* the little model as practice, and its solo writing did measurably improve (it even learned to actually *finish* its stories, which it rarely did before). But that's just the beginning. The frontier I'm heading toward is a proper **feedback curriculum**: not a teacher that corrects the same mistake forever, but one that turns each correction into a lesson the writer keeps — the way a child eventually stops needing to be asked "why did that happen?" because they've learned to ask it themselves. Right now my little writer has a teacher. Next, I want to give it an education. More soon.