A prompt is just the text you type to an AI. That is it. But it is also the single biggest lever on what you get back. Same model, two prompts, wildly different answers. The model is the engine, the prompt is the steering wheel, and most people are driving with their knees.

Before anything else, meet the unit that runs this whole world: the token. A token is a small chunk of text, roughly a word or a piece of one. The model reads your prompt in tokens and writes its answer in tokens. "Cat" is one token. "Unbelievable" might be three. Every token in and every token out is work the machine has to do.

That work has a name: inference. It runs on servers that pull electricity, and a lot of data centers use water to keep that hardware cool. So the chain is simple: more tokens means more inference, which means more energy and, yes, more water. Your lazy "uhh just write something about dogs" is not free. It just sends the bill somewhere you cannot see.

One honest note so no one can call us out later. You have probably seen scary stats like "one prompt equals half a bottle of water." Most of those numbers are shaky, averaged across very different data centers, or just old. We are not going to pretend to a precision nobody actually has. The mechanism is real and that is enough: fewer wasted tokens, less wasted compute. Full stop.