Don't just read about AI. Poke it. Play with it. Beat it.
The home of @ai.4.beginners — AI news explained simply, plus tiny experiments you can poke: guess against a language model, crank a chaos dial, shred a sentence into tokens. You'll understand AI because you touched it.
A language model writes by predicting the next word. Think you can predict it too? Five rounds.
winner gets bragging rights →Which word does the model pick next?
“Temperature” controls how adventurous an AI gets. Low = predictable. High = unhinged poetry. The bars are computed live — this is the real math.
drag it all the way, we dare you →“Describe a cat.” →
Models don't read words — they read tokens, chunks about ¾ of a word long. Type anything. Watch it get shredded.
try your name! →≈ 0 tokens ·
AI news, explained like you're human.
One author, no guest posts — the same stories as @ai.4.beginners on Instagram, with a little more room to breathe.
Eight words that unlock every AI article.
A system trained on vast amounts of text to predict the next word. The thing you just beat (or didn't) in Game No. 1.
The request you give a model. The single biggest lever on answer quality you control — master it and every AI tool gets better.
A chunk of text, roughly three-quarters of a word — the unit a model actually reads and writes in. You shredded some in Toy No. 3.
The dial that controls how adventurous a model's word choices are. You turned it in Experiment No. 2. Handle with care above 1.5.
A confident, fluent, entirely made-up answer. Not a lie — a plausible guess presented as fact. See: the blog post about fake sources.
The text, images, and examples a model learned from. It shapes — and limits — everything the model can do.
Layers of simple math, loosely inspired by neurons, that turn examples into ability. The machinery under every modern model.
An AI that doesn't just answer but acts — browsing, writing files, using tools in sequence to finish a task.