STEM
TokenTales
Ever wonder how AI "writes"? Train a tiny model on a nursery rhyme, then predict what it will say next — and discover the secret: a language model just counts which word tends to follow which, then guesses the most common one. TokenTales builds a real (tiny) next-word model right in your browser — no chatbot, no internet, nothing leaves your device — so you can see exactly how the trick works, and even watch the model write. Ages 9–14.
#6d5bd0 Distributed-narrative cast
Meet the cast
TokenTales' crew ARE the pieces of a language model — the cast IS the idea: Token is one word (a model reads and predicts one at a time), Tally is the counter (how often each word followed another — the model's whole memory), Gram is the look-back (a bigram uses the last 1 word, a trigram the last 2). Mentor Nova keeps the honest point: the model has no idea what words MEAN — it predicts the word that most often came next; knowing that is knowing how AI really works. Painted portraits + storybooks are a tracked follow-on (R-WEB-CLONE-SPAWN-TRACKED); a text-only crew + a flat, deterministic on-device model ship now — no LLM, nothing leaves the device.
Token
A single word (or piece of a word) — a language model reads and predicts one token at a time
Tally
The counter — how many times each word followed another; the counts ARE the model's memory, nothing more
Gram
The look-back — a bigram uses the last 1 word, a trigram the last 2; more look-back = smoother writing
Nova
(Mentor) — the model has no idea what words MEAN; it predicts the most common next word. Knowing that is knowing how AI really works.
What's inside
Learning goal
Ever wonder how AI "writes"? Train a tiny model on a nursery rhyme, then predict what it will say next — and discover the secret: a language model just counts which word tends to follow which, then guesses the most common one. TokenTales builds a real (tiny) next-word model right in your browser — no chatbot, no internet, nothing leaves your device — so you can see exactly how the trick works, and even watch the model write. Ages 9–14.
Question kits
16 curriculum-aligned kits × 25 questions = 400 questions per app, mapped to recognized standards.
On-device AI mentor
FoundationModels-powered hints, feedback, and adaptive difficulty — all running locally.
Mentored by Nova — on-device AI, no data leaves the device.
How TokenTales handles your kid's data
- ✅ All progress, settings, and AI-generated content stays on the device
- ✅ No analytics, no tracking, no third-party SDKs
- ✅ No ads, no in-app purchases — you pay once
- ✅ COPPA compliant under the 2026 FTC amendments
- ✅ Parental controls + session limits + content filters built in
TokenTales runs on ForgeKit — the open-source Swift Package Manager framework that powers every Spark & Anvil app. ForgeKit ensures consistent accessibility, COPPA compliance, and design language across the portfolio, so your kid's progress and preferences feel coherent across every app they touch.
Coming to the App Store
TokenTales is in active development. Email us to hear when it ships — no marketing, no spam, just a one-shot launch announcement.
Email me at launch