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

The token cascade, explained

Three things you need to know: the four pillars (what tokens are), the cascade (how they compound), and the yield formula (how we measure it). Visual, not abstract.

The four pillars

Every turn, tokens flow through the model in four ways. Two go in (fresh input + reused cache). Two come out (the answer + saved context). The ratio between them is your cascade.

Inputfresh tokens · $3/MCache Readreused context · $0.30/MMODELprocesses tokensOutputthe answer · $15/MCache Writesaved context · $3.75/Mentersreusedgeneratessavesnext turn: saved context becomes reusable context↓ to you

Cache read costs $0.30/M — 10× cheaper than fresh input at $3/M. The cascade is the art of replacing input with cache read.

The cascade

Each turn, output becomes cache write, which becomes cache read for the next turn. Fresh input shrinks. Cache grows. Output compounds. This is the snowball.

Turn 1Input-BoundICRΥ=0.05Turn 5Deep ReaderICRΥ=1.5Turn 15ArchivistICRΥ=3.54Turn 50Convergent ★ICRΥ=257.3InputCache ReadCache WriteOutput

By turn 50, fresh input is 2 tokens. Cache read is 2.5 million. The model already knows the whole project. You stopped re-explaining. The cascade compounds on its own.

The yield formula

Yield (Υ) is the headline number. It measures how much your cascade compounds. Two factors: how much you reuse (leverage) and how much you produce (velocity).

velocity= Output ÷ InputIOleverage= Cache Read ÷ InputICR×Υ Yieldleverage × velocityyour rankWhere you sit:Input-Bound0.05Deep Reader1.5Archivist3.5Amplifier26.5Convergent ★257.3
velocity
O ÷ I
output per fresh input
leverage
CR ÷ I
cache reuse per fresh input
yield
lev × vel
the headline number
field median
1.24
the human center of mass

Ready to see your own cascade? Get scored →