Early access — cascade metrics are real (derived from canonical token telemetry); the operator field is a curated seed. Learn more about the data
◈ Yield Calculator

Yield (Υ) Calculator

Enter your pillars for Υ Yield and see your Υ Yield score and cascade shape instantly. Pure arithmetic — nothing leaves your browser.

Enter your four token pillars

Pull these from ccusage --json or npx sigrank me. Token counts only — no prompt content.

Υ Yield2.50
Compression: 0.38Cache hit: 84%Leverage: 6.7x
High yield

A productive cascade — good cache reuse and solid output per input. Compounding, not just burning.

Yield thresholds are approximate (low < 0.5, moderate 0.5–2, high 2–10, very high 10+). Class tier is based on total tokens, not yield — see the class checker. For the full paste-based scoring experience with signed submission, see /score.

How yield is computed

Υ = cache_read × output / input². Yield rewards two things at once: reusing cached context (cache_read) and producing dense output per fresh token (output / input). An operator who re-sends the same large context every turn burns input without compounding — low yield. An operator who caches aggressively and extracts high signal per fresh token compounds — high yield. Volume alone is noise; yield is signal.

The four pillars map directly to the metrics most AI platforms now expose: input (fresh prompt tokens), output (generated tokens), cache-read (prompt-caching hits), and cache-write (new tokens written to cache for future reuse).

Frequently asked questions

What is the Yield (Υ) metric?

Yield (Υ) = cache_read × output / input². It measures the architecture of an AI token cascade — whether cached context is compounding signal into output, or tokens are being burned. It is the headline SigRank operator-efficiency metric.

How is the yield calculator different from the /score page?

This calculator takes four raw token counts and shows the math instantly in your browser. The /score page is the full paste-based flow that can also submit a signed snapshot to the leaderboard. Use this for a quick estimate; use /score to compete.

Where do I get my token counts?

Run `ccusage --json` or `npx sigrank me` to read the four token pillars (input, output, cache-read, cache-write) from your local AI session logs on-device. No prompt content is ever read — only the four integers.

What is a good yield score?

Roughly: below 0.5 is low yield (just starting), 0.5–2 is moderate yield, 2–10 is high yield (a productive compounding cascade), and 10+ is very high yield (signal compounds aggressively). These thresholds are approximate. Class tier is based on total tokens, not yield — see the class checker tool.

Does the yield calculator send my data anywhere?

No. All arithmetic runs in your browser. Nothing is transmitted, stored, or logged. The calculator is a pure client-side estimation tool.

Related: Yield (Υ) · Class Checker · Improve Your Yield