8 min · +70 XP
What AI Actually Is
A clean mental model you'll never have to unlearn.
By the end of this lesson
You'll be able to explain modern AI to a friend in one sentence.
Forget robots and forget sci-fi. The kind of AI that changed everything recently has a boring, beautiful core idea: it predicts what comes next.
Your phone guesses your next word from the last few. Now imagine an autocomplete that read a meaningful slice of everything humans ever wrote, and guesses not the next word but the next paragraph, the next plan, the next line of working code. That's a large language model.
That sounds too simple to be powerful. But 'predict the next piece, brilliantly, having absorbed humanity's writing' turns out to be enough to write, explain, plan, translate, code, and reason. Power emerged from scale.
Modern AI is a system that learned the patterns in human language and knowledge so well that it can continue almost any task you start — in words, code, or ideas.
It is not conscious. It is not a database it 'looks things up' in. It is not always right. It is a pattern engine — astonishingly useful, occasionally confidently wrong. Treat it like a brilliant, eager, slightly overconfident collaborator. (And when your chat AI 'searches the web' with sources? That's a tool bolted onto the predictor. Claude Code uses tools the same way — the core is still prediction.)
What is a large language model, at its core, actually doing?
Author
Evgeny Arsentyev
PhD · Chief Product Officer at a healthtech company