Spend Optimization vs Operator Scoring

The short version

CodeBurn is a free, local-first tool that tracks AI coding token usage and cost across 41 tools and agents. It goes beyond cost tracking: it scans for waste patterns (re-read files, low read:edit ratio, unused MCP servers, bloated CLAUDE.md), can apply fixes automatically with undo, guards your budget with session caps, compares models on one-shot rate and cost per edit, and correlates sessions to git commits to see whether spend actually shipped. It is a serious spend-optimization tool.

SigRank reads the same token telemetry and asks a different question: is your cascade compounding or burning? The headline metric, Υ Yield = cache_read × output / input², measures the architecture of your token cascade. SigRank derives compression ratio, SNR, Leverage, Velocity, and 10xDEV, assigns class tiers, and publishes ed25519-signed snapshots to a public leaderboard. CodeBurn optimizes your setup; SigRank scores your skill. Both matter. They are not substitutes.

Feature comparison

Two “yield” concepts, different meanings

Both tools use the word “yield,” but they measure different things. CodeBurn’s yield correlates AI sessions to git commits — it classifies spend as productive, reverted, abandoned, or ambiguous based on whether commits landed in main. It answers “did the money I spent actually ship?”

SigRank’s Υ Yield (cache_read × output / input²) measures the architecture of your token cascade — whether cached context is compounding into output or fresh input is burning without leverage. It answers “am I driving my AI tools efficiently?” One is a git-attribution metric; the other is a token-efficiency metric. Both are useful; they are not the same thing.

When to use which

Frequently asked questions

Is SigRank a CodeBurn alternative?

They overlap on reading AI coding token logs but answer different questions. CodeBurn optimizes your spend — it scans for waste patterns (re-read files, low read:edit ratio, unused MCP servers, bloated CLAUDE.md), can apply fixes automatically, guards your budget with session caps, compares models, and correlates sessions to git commits. SigRank scores your cascade efficiency (Υ Yield) and ranks you on a public leaderboard against other operators. If you want to reduce your AI bill, CodeBurn is built for that. If you want to know how efficiently you drive your AI tools compared to everyone else, SigRank answers that. You can run both — they read the same logs.

CodeBurn has a “yield” feature. How is it different from SigRank’s Yield?

They use the same word for different things. CodeBurn’s yield correlates AI sessions to git commits — it classifies spend as productive, reverted, abandoned, or ambiguous based on whether commits landed in main. It answers “did the money I spent actually ship?” SigRank’s Yield (Υ = cache_read × output / input²) measures the architecture of your token cascade — whether cached context is compounding into output or fresh input is burning without leverage. It answers “am I driving my AI tools efficiently?” One is a git-attribution metric; the other is a token-efficiency metric. Both are useful; they are not the same thing.

What does CodeBurn do that SigRank doesn’t?

CodeBurn does several things SigRank doesn’t: (1) waste scanning — it identifies specific inefficiencies like re-read files, low read:edit ratios, unused MCP servers, and bloated CLAUDE.md files; (2) fix application — it can apply fixes to your config interactively, with undo and auto-revert; (3) budget guarding — it installs hooks into Claude Code that warn or stop sessions at configurable cost caps; (4) model comparison — it compares models on one-shot rate, retry rate, cost per edit, and cache hit rate; (5) git-linked yield — it attributes sessions to commits to classify spend as productive or abandoned. SigRank doesn’t modify your config, guard your budget, or compare models. It scores the operator.

What does SigRank do that CodeBurn doesn’t?

SigRank scores and ranks operators. The headline metric, Υ Yield = cache_read × output / input², measures cascade architecture — whether signal is compounding or burning. SigRank derives compression ratio, SNR, Leverage, Velocity, and 10xDEV from the same four token pillars. It assigns class tiers (IGNITER to ARCH+), publishes ed25519-signed snapshots to a public leaderboard, and supports head-to-head operator comparisons. CodeBurn optimizes your local setup; SigRank measures where you stand against the field. If you want to improve your spending efficiency, CodeBurn is the tool. If you want to know whether your cascade architecture is competitive, SigRank is the tool.

Can I use both CodeBurn and SigRank?

Yes, and they complement each other well. Run CodeBurn to find and fix waste in your setup, guard your budget, and see whether your spend actually ships. Run SigRank to score your cascade efficiency and see where you rank. Both read the same local token logs — CodeBurn reads session files from 41 tools, SigRank’s scanner reads token counts on-device. Install SigRank with ‘npm install -g sigrank’, enroll, and submit a signed snapshot to the leaderboard. Keep CodeBurn for the budget and waste view.