TRANSPARENCY BEFORE YOU SEND

Should (A)I?

See the real cost of a prompt before you hit send — the dollars, the water, the energy. No judgment. Just the facts, and an honest estimate.

DRAWN BYCameron Beebe
SHOPStillwater Studio
SHEET06 · WEB TOOL
STATUSNo AI calls
WORK IN PROGRESS

Bucket ranges, pricing, and environmental constants are estimates and still being tuned — not guarantees. If a result looks off, send the prompt that caused it and I’ll take a look.

CHECK A PROMPT

Type it. See the range. Rework if you want.

MODEL
list prices · verified Jul 2026
YOUR PROMPT
0 words · 0 tokens tokenized locally

Measures your prompt only. Real chat apps also bill a hidden system prompt and conversation history — often 1,500–3,000+ tokens per turn — which isn’t counted here.

Enter a prompt and hit Check cost to see the breakdown.

The trust model

TWO KINDS OF NUMBERS

What we know versus what we estimate — drawn differently, on purpose. A blueprint convention borrowed on purpose: solid lines are drawn, dashed lines are projected.

EXACT · FACTS

Your input is measurable — counted locally and multiplied by public price-per-token. Drawn as a solid line.

~ESTIMATED · INFORMED GUESS

The response hasn’t happened yet — bucketed by intent and length, always shown as a range. Drawn as a dashed line, never a false-precise number.

Energy/water ranges are grounded in Google's Aug 2025 inference-impact paper, first-party OpenAI disclosures, and independent benchmarks (Jegham et al. 2025; UC Riverside, 2023) — not invented constants. Low bound = first-party, on-site-only figures; high bound = holistic estimates including indirect water and reasoning-heavy generation. Verified Jul 2026; efficiency figures move fast, so treat anything older than ~12 months as stale.