Token Cascade — The Flow of Tokens Through AI Coding Sessions

What a token cascade is

A token cascade is the flow of tokens through an AI coding session. Every turn moves tokens through four stages: you send fresh input to the model, the model generates output back, cache-read tokens are reused from prior context via prompt caching, and cache-write tokens are written to cache for future reuse. The cascade is the full picture of how tokens enter, circulate, and leave the session.

The metaphor is deliberate: tokens cascade like water through a series of pools. Some pools compound — cached context flows back into the next turn, amplifying a small fresh input into large output. Other pools drain — fresh input pours in, nothing is reused, and thin output trickles out. The cascade is the architecture that determines which one you have.

The four pillars of the cascade

Every token in an AI coding session falls into exactly one of four buckets. Together they describe the full flow of the cascade.

Token cascade vs raw token consumption

Token consumption is a single number: total tokens used. It tells you how much you spent. Token cascade is the architecture: how those tokens flowed through the four pillars. It tells you whether the spend compounded or burned.

Two operators can consume the same 500K tokens and get wildly different results. One reuses cached context efficiently — high cache-read, low input, high output — and her yield is high. The other sends fresh prompts without reusing context — high input, low cache-read, low output — and his yield is low. Same volume, different cascade, different skill. The yield metric Υ = cache_read × output / input² captures this in one number.

This is why cascade architecture — not raw volume — is the measure of AI coding efficiency. Counting tokens tells you what you spent. Reading the cascade tells you how efficiently you spent it.

Why the cascade matters

The cascade matters because it is the unit of skill in AI coding. Your token cascade is your skill signature — the architecture of how you drive your AI tools. A compounding cascade (high cache reuse, high output per input) is the signature of an operator who structures turns to recall cached context and sends minimal fresh input. A burning cascade (high input, low cache, low output) is the signature of an operator who pays full price every turn.

The cascade is also privacy-preserving by construction. Four integers — input, output, cache_read, cache_write — fully describe the architecture without revealing a single word of what you typed or what the model returned. This is what makes a global, continuous operator ranking possible without reading anyone's prompts.

Explore the cascade

FAQ

What is a token cascade?

The flow of tokens through an AI coding session across four stages: input, output, cache-read, and cache-write. The cascade is the full picture of how tokens enter, circulate, and leave a session.

What are the four pillars?

Input (fresh tokens sent), output (tokens generated back), cache-read (cached tokens reused), and cache-write (new tokens written to cache for future reuse).

Cascade vs consumption?

Consumption is a single number — total tokens used. Cascade is the architecture — how those tokens flowed. Two operators with the same consumption can have very different cascades and very different yield.

How do I measure my cascade?

Install the SigRank CLI (npm install -g sigrank), run sigrank enroll, and submit a snapshot. The on-device scanner reads token counts locally and publishes a signed snapshot. No message content leaves your machine.