Using AI without the waste

You can’t control how a model was trained, but the way you use it does affect how much computing — and therefore energy — gets spent on your behalf. The good news is that the efficient habits are mostly the same as the effective ones: less waste, better answers.

Habits that help

• Ask once, well. A clear, specific prompt that lands first time beats ten vague follow-ups. Good prompting is also the low-energy path.
• Right-size the tool. Not every task needs the largest, most powerful model. A smaller model is cheaper and lighter for simple jobs.
• Don’t paste the whole book. Sending huge amounts of irrelevant text burns through the context window for no benefit. Give it the relevant part.
• Reuse instead of regenerating. Save a good output rather than re-asking from scratch each time.
• Batch related questions in one focused session rather than scattering them.

Keep it in proportion

None of this should tip into anxiety over a single chat — the per-use cost is small. It’s about not being needlessly wasteful, the same way you’d not leave the tap running.

The most efficient prompt is usually the best one: specific, scoped, and asked once. Saving energy and getting good answers point the same direction.

For the bigger picture on where the cost actually sits, see what AI costs the planet.