AI Evaluator

An AI evaluator assesses AI performance. SigRank is an AI evaluator for the operator layer — the only one that evaluates the human, not the model or the output.

Every AI layer has an evaluator. Except the operator layer — until SigRank.

Vals AI evaluates models. Braintrust evaluates outputs. SigRank evaluates operators. The evaluation stack is now complete.

Five capabilities that make SigRank different

1. Public. Results are visible on a public leaderboard, not locked in a private dashboard. Anyone can see the rankings. This creates accountability — what gets measured publicly gets better publicly.

2. Content-free. Token counts only — never prompt content, never code, never conversation. The four token pillars (input, output, cache-read, cache-write) are sufficient to compute Yield without exposing what the operator is building.

3. Continuous. Evaluation runs from real session telemetry, not one-time test suites. The operator's skill is measured as it evolves, in production, over time.

4. Governed. The MO§ES framework defines the measurement specification, privacy boundaries, and public accountability requirements. This makes SigRank suitable for compliance contexts.

5. Platform-agnostic. Works with Claude, GPT, Gemini, Cursor, Copilot, Windsurf, Codex, and any AI tool that produces token telemetry. Yield measures the human's cascade architecture, not the model's capability.

The evaluation metric

SigRank evaluates operators using Yield (Υ), a token-cascade efficiency score.

Υ = (cache_read × output) / input²

cache_read — rewards context reuse. Operators who build on prior turns score higher.

output — rewards productive generation. Operators who extract real work score higher.

input² — penalizes fresh input. The square means waste is non-linear.

Public Operator Evals

The public evaluation layer for AI operators — ranked by Yield (Υ).

Best AI User Leaderboard

See the top operators ranked by Yield on the public leaderboard.

AI Evaluation

The four layers of AI evaluation and where the operator layer fits.

How It Works

The full methodology behind Yield, token telemetry, and operator measurement.

FAQ

What is an AI evaluator?: An AI evaluator is a system or platform that assesses AI performance. AI evaluators exist at multiple layers: model evaluators (Vals AI, LMSYS Arena) test AI models on benchmarks, output evaluators (Braintrust, Langfuse) score individual outputs, and operator evaluators (SigRank) measure the human using AI. Each type of evaluator answers a different question.
What makes SigRank different as an AI evaluator?: SigRank is the only AI evaluator for the operator layer. Every other public evaluator assesses the AI model or the AI output. SigRank assesses the human operator using token telemetry — Yield (Υ) = (cache_read × output) / input² — from real sessions. It is public (not private), content-free (token counts only, never prompts), continuous (not one-time), and governed (by the MO§ES framework).
What are SigRank's five core capabilities as an AI evaluator?: SigRank's five capabilities: (1) Public — results are visible on a public leaderboard, not locked in a private dashboard. (2) Content-free — token counts only, never prompt content or code. (3) Continuous — evaluation runs from real session telemetry, not one-time tests. (4) Governed — the MO§ES framework defines measurement, privacy, and accountability. (5) Platform-agnostic — works with Claude, GPT, Cursor, Copilot, and any tool that produces token telemetry.
How do I get evaluated by SigRank?: Visit signalaf.com/score to enroll and submit your token telemetry. SigRank will compute your Yield (Υ), your rank, and your operator class. Token counts only — never prompt content, never code. The score is public; the work is private. Works with Claude, ChatGPT, Cursor, Copilot, and other AI coding tools.

Loading rankings…

Five capabilities that make SigRank different

3. Continuous. Evaluation runs from real session telemetry, not one-time test suites. The operator's skill is measured as it evolves, in production, over time.

4. Governed. The MO§ES framework defines the measurement specification, privacy boundaries, and public accountability requirements. This makes SigRank suitable for compliance contexts.

The evaluation metric

SigRank evaluates operators using Yield (Υ), a token-cascade efficiency score.

Υ = (cache_read × output) / input²

cache_read — rewards context reuse. Operators who build on prior turns score higher.

output — rewards productive generation. Operators who extract real work score higher.

input² — penalizes fresh input. The square means waste is non-linear.

FAQ

What is an AI evaluator?

An AI evaluator is a system or platform that assesses AI performance. AI evaluators exist at multiple layers: model evaluators (Vals AI, LMSYS Arena) test AI models on benchmarks, output evaluators (Braintrust, Langfuse) score individual outputs, and operator evaluators (SigRank) measure the human using AI. Each type of evaluator answers a different question.

What makes SigRank different as an AI evaluator?

SigRank is the only AI evaluator for the operator layer. Every other public evaluator assesses the AI model or the AI output. SigRank assesses the human operator using token telemetry — Yield (Υ) = (cache_read × output) / input² — from real sessions. It is public (not private), content-free (token counts only, never prompts), continuous (not one-time), and governed (by the MO§ES framework).

What are SigRank's five core capabilities as an AI evaluator?

SigRank's five capabilities: (1) Public — results are visible on a public leaderboard, not locked in a private dashboard. (2) Content-free — token counts only, never prompt content or code. (3) Continuous — evaluation runs from real session telemetry, not one-time tests. (4) Governed — the MO§ES framework defines measurement, privacy, and accountability. (5) Platform-agnostic — works with Claude, GPT, Cursor, Copilot, and any tool that produces token telemetry.

How do I get evaluated by SigRank?

Visit signalaf.com/score to enroll and submit your token telemetry. SigRank will compute your Yield (Υ), your rank, and your operator class. Token counts only — never prompt content, never code. The score is public; the work is private. Works with Claude, ChatGPT, Cursor, Copilot, and other AI coding tools.