Prompt You Can Paste Into Any AI This is a ready-to-use prompt for creating a small, shareable tool that tracks how often you use an AI assistant and estimates what it would cost to offset the energy, water, and carbon behind that use. How to use it Copy everything inside the box below. Paste it into any capable AI (ChatGPT, Claude, Gemini, Copilot, etc.). Before sending, replace the two bracketed placeholders: [YOUR BRAND NAME] and [ACCENT HEX] (a brand color; defaults to a craft red if you leave it). Save the HTML file the AI returns as usage-ledger.html and open it in any browser. It logs from that point forward and remembers your totals. Want a Windows desktop version instead of a browser file? Add one line to the prompt: "Build it as a single PowerShell + WinForms script instead of HTML." You are helping me build a personal "AI Usage Ledger" — a small, shareable tool that lets me log my interactions with an AI assistant and estimate what it would cost to offset their environmental footprint (energy, water, and carbon). Produce TWO things: (1) a single self-contained HTML file, and (2) a short methodology summary. HONESTY CONSTRAINTS (respect these and surface them in the output): - This is a SELF-REPORTED ESTIMATE, not a measurement. Most AI providers do not publish per-query energy/water/carbon figures for their specific models, so every footprint number is an estimate anchored to disclosed peer data and published studies, and could be off by several times. Say this plainly in both the tool and the summary. - Do not imply precision. Make clear that an individual's footprint is a tiny fraction of one purchasable offset unit. DELIVERABLE 1 — THE WIDGET (one HTML file, no external libraries, fonts, or network calls; everything inline; works offline in any modern browser): Features: - Three logging buttons for ask size: LIGHT (answering a question), MEDIUM (drafting text), HEAVY (creating a document). Each click logs one interaction of that tier. - A large running total plus a per-tier breakdown (light / medium / heavy counts). - An UNDO button that removes the most recent logged interaction, whichever tier it was. - A live readout that converts the totals into estimated energy (kWh), water (US gallons), and carbon (kg CO2e), and into offset purchase quantities and cost. - A "Copy summary" button that copies a plain-text summary to the clipboard. - Persist all data locally in the browser with localStorage so the count survives across sessions (tracking forward from first use). Also remember the window/section state if relevant. - Compact, branded, legible styling. Use accent color [ACCENT HEX, default #C04148] and brand name "[YOUR BRAND NAME]". Lead with a clear hero number for the total. - Accessible: keyboard-operable controls, visible focus outlines, and respect prefers-reduced-motion. FOOTPRINT MODEL — per interaction. Use exactly these defaults and expose them as clearly labeled editable constants near the top of the code so they can be tuned: Tier Energy (Wh) Water (mL) Carbon (g CO2e) Light (question) 0.5 5 0.5 Medium (drafting) 2 20 2 Heavy (document) 10 100 10 These keep a consistent ratio of about 10 mL of water and 1 g CO2e per watt-hour, scaled by ask size. For context (state this in the summary): disclosed figures put a typical text query near 0.24–0.34 Wh and about 0.3 mL of on-site water; a longer ~400-token reply has been measured in a life-cycle assessment at roughly 45 mL and 1.14 g CO2e; and heavy reasoning or long document generation can run 7–40 Wh. CONVERSIONS: - Energy total (kWh) = sum(count_tier x Wh_tier) / 1000 - Water total (liters) = sum(count_tier x mL_tier) / 1000; US gallons = liters / 3.78541 - Carbon total (kg) = sum(count_tier x g_tier) / 1000 OFFSET INSTRUMENTS AND REFERENCE PRICES (expose as editable constants): - Carbon credits: $20 per tonne CO2e (high-integrity nature-based; note the market spans under $5 for low quality to $150–$500+ for engineered removals). - Renewable Energy Certificates (RECs): $5 per MWh (voluntary market roughly $0.50–$10). - Water Restoration Certificates (WRCs): $2 per 1,000 US gallons restored. Offset quantities and cost: - Carbon credits (tonnes) = carbon_kg / 1000; cost = tonnes x $20 - RECs (MWh) = energy_kWh / 1000; cost = MWh x $5 - WRCs (units of 1,000 gal) = water_gallons / 1000; cost = units x $2 - Marginal total = sum of the three costs. REALITY-CHECK LINE to display in the tool: offsets sell in whole units (1 tonne CO2e, 1 MWh, 1,000 gallons), so an individual's footprint is a tiny fraction of one unit. Show the "smallest real basket" (one of each = $27 at the default prices) and how many times it over-covers the current footprint. DELIVERABLE 2 — THE METHODOLOGY SUMMARY (plain text or markdown, about half a page): - One short paragraph on what the tool is and isn't (self-reported estimate; no official per-model figures published; wide uncertainty). - The per-tier assumption table and the offset prices, with the context anchors above. - The three conversion formulas. - The minimum-tradable-unit caveat. - One line on how to act on it: either accept the over-coverage and buy one minimum unit of each, or pool many people or many years together so the purchase maps to whole units. OUTPUT ORDER: Give the complete HTML file first (ready to save as usage-ledger.html and open in a browser), then the methodology summary. No external dependencies of any kind. Customize before sharing Swap [YOUR BRAND NAME] and [ACCENT HEX] for each reader's brand. The defaults render a clean, on-brand widget if left as-is. The per-tier numbers and offset prices are intentionally exposed as editable constants in whatever the AI generates, so anyone can tune them as better figures become available. A note to include when you publish this The figures are deliberate, defensible estimates — not measurements — because per-query environmental data for specific AI models is mostly unpublished. Treat the output as order-of-magnitude, and revisit the constants as providers disclose more.