Pick the next word yourself, or let it auto-write. This is the core loop behind every AI writing tool, slowed down so you can see it happen. No login, nothing saved.
What's been written so far
What the AI thinks comes next — click one to add it
It predicts, it doesn't "know"
At its core, the model is estimating: given everything so far, what word is most likely to come next? Repeat that thousands of times and you get a full response.
"Temperature" controls the risk
Low predictability means safe, expected wording. Higher predictability trades some reliability for more varied, creative-sounding output.
Scale is the real difference
This toy version learned from a few paragraphs. Real models learn from a meaningful slice of the internet — same idea, vastly more capable.
What is a transformer?
The real architecture behind ChatGPT, Claude, and Gemini, explained plainly.
How AI image generators work
The same generative idea, applied to pictures instead of words.
Try the Tokenizer Playground →
See the actual chunks of text an AI model reads and writes.
Try Machine Learning Types →
See the training side of the story, not just the writing.
It means the AI creates new content rather than just sorting or labeling existing content. Text, images, audio, and video tools are all "generative" if they produce something new based on a prompt, instead of picking from a fixed list of pre-written answers.
Conceptually yes, though the real thing is far more sophisticated. This tool predicts the next word using simple word-pair patterns from a small sample of text. A real model like GPT or Claude predicts using patterns learned from a huge portion of the internet, weighing the entire conversation so far, not just the last word.
AI companies call this "temperature." Turned down, the AI always picks its single most likely next word, which produces safe, repetitive text. Turned up, it sometimes picks a less obvious but still plausible word, which produces more varied and creative — but occasionally odd — results.
This demo only knows the handful of sentences it was built from, so it can only continue from words it has actually seen used together before. A real AI model has effectively seen almost every common word combination in its training data, so this rarely happens in practice.
No. Everything runs instantly in your browser using a small, simplified word-prediction model — no API calls, no account, nothing sent anywhere. It's built to make the underlying idea visible, not to be a production AI writing tool.
Tokenizer Playground
Type anything and watch it split into tokens live — the units LLMs actually read and get billed for.
Context Window Visualizer
See how much of a model's context window your text fills up, and what happens when you run out of room.
FROG in a Bowl Prompt Builder
Fill in Format, Role, Objective, Goal, and Context — get a copy-ready structured prompt in seconds.
Machine Learning Types
Three tiny games show what supervised, unsupervised, and reinforcement learning actually mean.
Neural Network Playground
Drag two sliders and watch a real, tiny neural network turn them into a decision, live.
RAG Playground
Ask a question, watch notes get retrieved, then see a grounded answer versus a hallucination.
Embedding Map
Click two words and see why similar meanings sit near each other — the idea behind vector search.
Attention Visualizer
Click a word and see which others a toy transformer looks at — including the classic “it” puzzle.
Prompt Injection Lab
Watch a pasted email try to override a support agent, then flip a switch that treats it as data.
