ScaleDown

We all slip into “email mode” when we chat with AI— lots of pleasantries, back-story, and polite filler. Large-language models don’t hear any of that; they just count tokens. Extra words raise API bills, slow responses, and keep GPUs burning a little longer than they need to. ScaleDown is a free, open browser extension that bridges that gap. As you type, it offers a second version of your prompt—stripped to the essential verbs, nouns, and context the model actually understands—while showing the tokens, cents, and grams of CO₂ you just saved. One click sends either version, so you keep full control and learn by comparison. Why people join the ScaleDown community - Cost-minded teams see 20-40 % fewer tokens per request, which compounds into real savings on volume workloads. - Prompt newcomers watch the rewrite in real time and pick up best practices without reading a single tutorial. - Climate advocates like that leaner prompts mean fewer compute cycles; the tool turns an abstract footprint into a visible, per-prompt number. - Power users & researchers share edge cases, benchmark data, and open-source snippets to push the ceiling on prompt efficiency. What you’ll find on this page - Bite-size experiments and before/after prompt examples - Shared dashboards of community token & carbon savings - Road-map polls—because the next feature should solve a problem you actually have ScaleDown started as a weekend project to trim our own OpenAI bill. It’s now a growing network of writers, analysts, engineers, teachers, and the simply curious—anyone who’d like to get more from every token and leave a lighter footprint while they’re at it. If that sounds useful, follow along, install when you’re ready (it’s one click, no sign-up), and tell us what you learn. The tool is free; the knowledge is shared; the impact scales with all of us.