Loading chapter
A short, honest answer. No jargon, no metaphysics. If you've never used an AI assistant before, start here.
Artificial intelligence - AI - is software that can recognize patterns in language and produce useful new text from them. It isn't a person. It doesn't think the way a person thinks. It is, roughly, a very large pattern-matching engine that has read a staggering amount of text and learned what tends to follow what.
If a calculator does what you tell it, an AI tries to do what you mean.
When you type gen: write a poem about cats, Typilot breaks the request into three parts: a verb (gen: = generate), a shape (poem), and a topic (cats). The model has seen thousands of poems and predicts what one about cats should look like.
The result isn't retrieved - it's composed, one token at a time, against the patterns the model learned during training.
It composes, it does not retrieve.
You use it when you'd otherwise:
The win is the round-trip - your fingers don't leave the keyboard, and the tool you need is already there.
Most AI tools send your text to a server. We don't. The model runs on your machine via Ollama, which means:
You trade some quality (the largest cloud models still beat what fits on a laptop) for full privacy and offline reliability. For most everyday writing and coding, the difference is small. For sensitive work, the difference is the whole point.
No. AI is a tool that helps you work faster, like a spreadsheet helped accountants. You still pick the question, judge the answer, and decide what ships.
No. You write what you want in plain language; the model handles the rest. The built-in commands cover most use cases.
Yes. The model is confident even when it's wrong. Review anything that matters before sending or shipping it.
Autocomplete suggests the next word from your own history. AI generates new text that fits the meaning of your prompt - phrases it has never seen verbatim.