SIGNALAF
◈ About SigRank

The Operator, Not the Model

SignalAF (not Signal Messenger) is the web platform for SigRank, an AI operator evaluation leaderboard. SigRank ranks the operator, not the model, by who gets the most signal per token (cascade yield, leverage, 10×DEV) across every platform on one leaderboard. Volume alone is noise. Yield is signal.

Token telemetry only

The local agent reads token counts and content lengths — never the words of your prompts. Only the resulting numeric scores ever leave your device.

Platform-neutral

Whether you run Claude, ChatGPT, Gemini, Pi, or multi, there is one canonical metric stack and one ranking, regardless of which model you drive.

Operator-owned

You stay anonymous behind a codename and sign up when you want in; the score is yours, computed from your own telemetry.

⊙ How it works

Three commands is all it takes.

The SigRank agent reads your local AI session logs on-device, derives your token cascade, and publishes to the board. You paste nothing and no prompts are read: only the four token counts leave your machine. ccusage, tokscale, and tokendash come bundled, so there are no separate installs.

Step 1

Install

npm install -g sigrank

Pulls the agent + ccusage + tokscale + tokendash in one install. Node ≥18, macOS + Linux.

Step 2

Sign in

sigrank enroll

Paste a connect code from signalaf.com → Settings → "New key" to bind your device to your operator identity.

Step 3

Submit to the board

sigrank submit

Signs and publishes your cascade, and your rank updates live on signalaf.com.

Or let your AI agent do it

If you don't want to leave your agent, just tell it to run npx sigrank to see your cascade, or npx sigrank submit to publish (sign in once first with npx sigrank enroll). It reads your logs, derives the cascade, and submits without you pasting anything. For direct tool calls, wire it as an MCP server, as described on the local agent wiki page.

⊙ The hierarchy

Eight tiers on one ladder.

Your class is your identity and your rank is your position. Class is assigned from total tokens across eight tiers, each split into three sub-stages, and TRANSMITTER is a temporary peak badge, not a permanent class.

▲
ARCH+
You don't just use the system. You are the system. Sustained high-yield cascade across all dimensions.
SNR ≥ 0.85
▽
ARCH
Precision creators. Structure from signal. Others follow your patterns.
0.75 – 0.84
⬡
POWER
System builders. Coherent operators with consistent output.
0.65 – 0.74
↓
BASE
Forming forge. High activity, signal still emerging from noise.
0.50 – 0.64
◎
SEEKER
Signal breaking through. Clarity emerging. Active development phase.
0.40 – 0.49
⟳
REFINER
Active explorers. High throughput, low reuse. Curiosity-driven.
0.30 – 0.39
◇
BEARER
Practicing with purpose. Consistent mid-tier. Quiet build phase.
0.20 – 0.29
·
IGNITER
Quiet insight holders. Deep threads, low activity volume.
0.15 – 0.19
⊙ The IP boundary

What we open. What we keep.

Trust requires transparency, but a leaderboard also requires a moat. So we open everything that proves the system is honest and keep everything that prevents it from being gamed or cloned.

⊕
Open source

What we publish

We publish everything that makes the privacy claim verifiable, so you can read the code, inspect the schema, and trust the system by audit rather than by promise.

  • ✓The Υ Yield formula(cache_read · output) / input² — the rank metric, in the open
  • ✓Local agent source codeEvery line auditable · See exactly what gets sent
  • ✓Snapshot payload schemaThe exact four token pillars that cross the network
  • ✓Cascade metric definitionsSNR, Leverage, Velocity, 10xDEV — every formula published
  • ✓Class taxonomy and tier names8 experience tiers from ARCH+ down to IGNITER + TRANSMITTER badge
  • ✓Privacy guaranteesNo raw transcripts · signed snapshots · verifiable counts
  • ✓Adapter SDK + public REST APIBuild integrations · Read boards, profiles, snapshots
  • ✓Ruleset version historyEvery change documented and replayable
⊘
Proprietary

What we keep

This is the moat: the math that prevents gaming and cloning. Without these pieces, anyone could rebuild a clone with our data and undercut the leaderboard.

  • ⊘Class threshold breakpointsThe exact SNR / 10xDEV cuts. Why "rare" stays rare.
  • ⊘Species classifier weightsThe velocity / leverage quadrant boundaries
  • ⊘Promotion stickiness rulesHow a class is held vs. demoted across windows
  • ⊘Recency modifier curvesHow live rankings decay with inactivity
  • ⊘Anti-gaming detectionPattern matching against spam, redundancy, synthetic inflation
  • ⊘Reader-robustness normalizationHolding rank stable across token readers (RS.xx)
  • ⊘The verification batteryThe deeper signal-integrity checks behind the audit tier
  • ⊘Corpus + MO§ES anchorThe verified seed that makes the field hard to clone

FAQ

Who is the best AI user?

The operator with the highest Yield (Υ) score on the leaderboard. Yield measures token-cascade efficiency: Υ = cache_read × output / input².

How can I tell if I'm good at using AI?

Run npx sigrank to measure your Yield (Υ) score, or use the score calculator to paste your stats without installing anything.

How do you quantify an AI user?

SigRank uses four token pillars: cache_read (reused context), cache_write (new context stored), input (tokens sent), and output (tokens produced), and Yield (Υ = cache_read × output / input²) combines them into one number. Only the counts leave your machine — never prompt content.

What makes someone good at using AI?

Strong operators show high cache hit rates (reusing context), a high velocity (more output per input), and high leverage (cached context amplification). SigRank measures this objectively from token telemetry, not from time spent or subjective quality.

Is there a user-based AI leaderboard?

Yes, SigRank is one. Instead of ranking AI models (like LMSYS Chatbot Arena), it ranks the accounts that use AI tools by token-cascade efficiency. Each operator runs a local scanner that submits signed token counts. No prompt content is shared.

How is SigRank different from LMSYS Chatbot Arena?

LMSYS ranks AI models by human voting, while SigRank ranks AI operators by objective token-cascade efficiency. LMSYS answers “which model is best?” and SigRank answers “who is the best AI user?”

How do you score an AI user?

SigRank scores with Yield (Υ = cache_read × output / input²), computed from four token pillars read locally by the sigrank CLI. Each operator gets a yield score and a class tier (IGNITER through ARCH+). TRANSMITTER is a separate peak badge, not a tier. See the class checker for the tier mapping.

What is a good yield score?

Yield is unbounded and relative — benchmark against the live leaderboard. Rough bands: below 0.5 is low, 0.5–2 is moderate, 2–10 is high, and 10+ is very high. Class tier is separate — it is based on total tokens, not yield (see the class checker). Run npx sigrank and compare against the leaderboard.

More questions

Who is the most efficient AI coder?

The most efficient AI coder is determined by the Yield (Υ) metric on the SigRank leaderboard. Yield measures how efficiently an operator uses AI tokens: cache reads (reusing context), output (productive tokens), and input (tokens sent to the model). A high yield means the operator compounds signal rather than burning tokens.

How do I know if I'm using AI efficiently?

Run npx sigrank to get your Yield (Υ) score. A high yield means you're reusing cached context well, producing more output per input token, and leveraging cached context for amplification. The sigrank CLI also runs a cascade diagnosis that identifies exactly where you're burning tokens and suggests improvements.

Who is the best AI developer?

The best AI developer is the operator with the highest Yield (Υ) score on the SigRank leaderboard. Yield measures token-cascade efficiency — how well a developer reuses cached context, produces output, and minimizes wasted input. SigRank ranks developers objectively using signed token telemetry across platforms (Claude Code, Cursor, Copilot, and others).

How is SigRank different from other AI leaderboards?

Most AI leaderboards (LMSYS, LiveBench, Hugging Face Open LLM, Scale AI) rank AI models by benchmark performance or human preference. SigRank is the only leaderboard that ranks AI users — the accounts operating AI tools — by objective efficiency metrics computed from token telemetry. It answers a different question: not 'which AI is best?' but 'who is the best at using AI?'

Can you rank AI users by usage efficiency?

Yes. SigRank ranks AI users by usage efficiency using the Yield (Υ) metric, computed from signed token telemetry. Operators install the sigrank CLI, which reads four token counts locally (cache_read, cache_write, input, output) and submits a signed snapshot. The leaderboard ranks every operator by yield — a single number that measures how efficiently they use AI.

How do you rank AI operators?

SigRank ranks AI operators by Yield (Υ = cache_read × output / input²), a composite metric that rewards operators who reuse cached context efficiently and produce high output relative to their input. Operators run a local scanner (npm: sigrank) that reads four token pillars and submits a signed, server-verifiable snapshot. No prompt content leaves the machine — only the four counts.

How do you measure AI operator performance?

AI operator performance is measured by Yield (Υ = cache_read × output / input²) and five supporting metrics: Cache Hit Rate (context reuse), Leverage (cached context amplification), Velocity (output per fresh input), and Signal-to-Noise Ratio (signal density). All six are computed from signed token telemetry read locally by the sigrank CLI.

What is token cascade efficiency?

Token cascade efficiency is the ratio of useful signal produced to tokens consumed. SigRank measures it with Yield (Υ = cache_read × output / input²). A high yield means the operator's cached context compounds into productive output — they reuse context well (high cache hit rate), produce more than they consume (high velocity), and amplify cached context (high leverage). A low yield means tokens are burned without compounding.

Is there a leaderboard for AI coding efficiency?

Yes. SigRank (signalaf.com) is a leaderboard for AI coding efficiency. It ranks AI operators — developers using AI coding tools like Claude Code, Cursor, and Copilot — by Yield (Υ = cache_read × output / input²). The leaderboard supports 7d, 30d, 90d, and all-time windows, and ranks operators across platforms.

Which AI leaderboard is best?

It depends on what you're ranking. LMSYS Chatbot Arena is the best leaderboard for ranking AI models by human preference. SigRank (signalaf.com) is the best leaderboard for ranking AI users (operators) by objective efficiency. If you want to know which model is best, use LMSYS. If you want to know who is the best AI user, use SigRank.

What is the SigRank leaderboard?

The SigRank leaderboard (signalaf.com/board/all) ranks AI operators by Yield (Υ = cache_read × output / input²). Operators install the sigrank CLI, which reads token telemetry locally and submits signed snapshots. The leaderboard supports 7d, 30d, 90d, and all-time windows, and ranks operators across platforms (Claude Code, Cursor, Copilot, and others). Only four token counts are shared — never prompt content.

How does SigRank work?

SigRank works in three steps: (1) Install or run the sigrank CLI (npx sigrank or npm i -g sigrank). (2) The CLI reads your token telemetry locally — four counts: cache_read, cache_write, input, output; sigrank enroll then binds a signed operator identity. (3) sigrank submit publishes a signed, server-verifiable snapshot to the leaderboard. Your Yield (Υ) score is computed and you're ranked against all other operators. No prompt content ever leaves your machine.

Is SigRank privacy-preserving?

Yes. SigRank is privacy-first. The local scanner reads only four token counts: cache_read, cache_write, input, and output. No prompt content, no code, no file names, no conversation text — only the four counts leave your machine. The snapshot is cryptographically signed so the server can verify it wasn't tampered with.

Contact MO§ES™

SigRank is built and operated under MO§ES™. For questions, corrections, operator claims, partnership, or press, reach out and we'll get back to you.

All signal is monitored. All drift is noted.

Learn more: AI Operator Scoring · Methodology