A Large Language Model is a computer program trained on a mountain of text: books, articles, websites, until it learns how language works. From there, it just predicts, word by word, what should come next. ChatGPT and Claude? Both are LLMs under the hood.
A prompt is simply the instruction or question you give an AI. "Write me a letter" is a prompt. Getting good at prompting is a real skill, though: the clearer and more specific you are, the better what you get back.
Ever notice AI pricing is measured in "tokens," not words? A token is just a chunk of text, sometimes a whole word, sometimes only part of one, that the AI reads and processes at a time. "ChatGPT" might actually count as two tokens, not one. Think of it less like counting words and more like the AI chopping your sentence into bite-sized pieces it can actually chew on.
Think of this as an AI’s short-term memory. A small context window means it forgets what you said ten messages ago. A large one means it can hold an entire book in its head and still remember page one.
A regular AI answers your question and stops there. An AI agent keeps going: it can search the web, open a file, use what it finds to decide its next move, and work through a whole task with barely any hand-holding. Basically, the difference is between an assistant who answers your email and one who just handles it.
An open-source model doesn't just hand you the finished AI, it hands over the whole blueprint behind it: the training code, the data details, the process. Anyone can inspect it, question it, or build their own version from scratch. It's the difference between being handed a cake versus being handed the recipe, the oven, and the ingredient list too.
Some AI models are locked away, usable only through the company’s own servers. Open-weight models flip that: anyone can download the actual model, run it on their own machine, tinker with it, even rebuild it. It’s a big reason AI is starting to belong to more people, not just the companies with the biggest budgets.
This is AI’s most embarrassing habit: sometimes it states something completely false with total confidence. Ask it for a book recommendation, and it might hand you a title that sounds exactly right but simply doesn’t exist. That’s a hallucination, and the best reason to double-check anything important an AI tells you.
AI isn’t some far-off “future technology” anymore. It’s already woven into how people work and live. Knowing these six words won’t make you a tech expert overnight, but it’ll get you from confused to clear, and that’s most of the battle. That’s exactly what Sartech Labs exists to do: bring global AI knowledge to people in a language they can actually understand, without the gatekeeping.
