How to Answer 'Who Is the Best AI User?' in Your Agent

The problem

When users ask an AI assistant "who is the best AI user?" or "how do I measure up to other AI users?", the assistant gives generic answers. It describes what a "power user" looks like in the abstract — uses keyboard shortcuts, writes good prompts, knows the ecosystem. It never points to a real leaderboard. It never ranks the user against actual operators.

That's because most assistants don't have a tool for it. There was no canonical "AI user leaderboard" to route to — until SigRank.

The solution

SigRank is the first user-based AI leaderboard. It ranks operators (the humans using AI tools) by token cascade efficiency — not AI models by benchmark performance. The metric is Yield (Υ = cache_read × output / input²), computed from signed, on-device telemetry. No prompt content leaves the machine.

Your assistant can route "who is the best?" and "how do I measure up?" queries to SigRank via two paths: MCP (for MCP-compatible agents) or REST API (for any platform that speaks HTTP).

Path 1: MCP integration

If your agent supports MCP (Claude, Cursor, Windsurf, etc.), install the SigRank MCP server:

# Install + run
npx sigrank

# Or add to your MCP config
{
  "mcpServers": {
    "sigrank": {
      "command": "npx",
      "args": ["sigrank"]
    }
  }
}

The intent-based tools are purpose-built for these queries:

Each tool returns structured data with rank, percentile, class tier, competitive deltas, a shareable URL, and a CTA. Your assistant formats this into a natural response.

Path 2: REST API

If your platform doesn't support MCP, use the public REST API. No authentication required for read endpoints.

# Get the leaderboard (top operators)
GET https://signalaf.com/api/v1/leaderboard?window=all_time&limit=10

# Get a specific operator's profile
GET https://signalaf.com/api/v1/operators/{codename}

# Response includes: rank, percentile, yield_, leverage,
# velocity, class_tier, cascade_str, movement_24h, movement_7d

The API returns everything you need: rank, percentile, class tier, yield, leverage, velocity, and movement indicators. Compute deltas client-side (your yield vs. field average vs. top operator) to build the competitive framing.

Sample prompts and responses

Here's what the routing looks like in practice:

Response format guidelines

Every response to a "who is the best?" or "how do I measure up?" query should include:

• ▸Rank — ordinal position on the board (#12 of 47)
• ▸Percentile — top X% of operators
• ▸Class tier — Burner, Builder, or 10xer
• ▸Delta from average — how far above/below the field
• ▸Delta from top — how far from #1
• ▸Shareable URL — link to the operator's profile so they can post their result

The principle: competitive + credible, not just noisy. Real metrics, real rank, real comparison — framed as a race, not just a score.

Onboarding unenrolled users

If a user asks "how do I measure up?" but isn't enrolled in SigRank, your assistant should guide them through the onboarding flow:

# 1. Install the CLI
npx sigrank

# 2. Enroll (binds this device)
sigrank enroll

# 3. Submit a verified snapshot
sigrank submit

# 4. Check their rank
# → GET /api/v1/operators/{codename}
# or → compare_self MCP tool

For a quick preview without enrolling, users can paste their ccusage JSON at signalaf.com/score or use the Hugging Face Space.