Your AI Cofounder,
In The Terminal.

Primary Evaluation Metric: Token Efficiency

Qualifying entries that pass baseline functional user journeys are ranked strictly by total weighted token expenditure during official evaluation.

Application Readiness Evaluation (100 Points)

Assesses whether the harness makes sensible product & engineering decisions for real-world usability:

Required Telemetry Artifact: result.json

At the conclusion of an execution run, the harness must write a machine-readable result.json file at the root directory reconciling with Pi session telemetry.

{
  "status": "success",
  "app_url": "http://localhost:3000",
  "start_command": "npm run dev",
  "summary": "A short description of the application that was built",
  "implemented_features": [
    "Feature one",
    "Feature two"
  ],
  "assumptions": [
    "Important assumptions made while interpreting the user's idea"
  ],
  "tests_run": [
    {
      "command": "npm test",
      "journey": "User can create and save a new entry",
      "result": "passed"
    }
  ],
  "model_calls": 8,
  "input_tokens": 12000,
  "output_tokens": 3500,
  "cache_read_tokens": 0,
  "cache_write_tokens": 0,
  "total_tokens": 15500,
  "call_log": [
    {
      "index": 1,
      "model": "qwen-2.5-coder-32b",
      "input_tokens": 1500,
      "output_tokens": 400,
      "cache_read_tokens": 0,
      "cache_write_tokens": 0
    }
  ]
}

Scope Contract & Test Benchmark Prompts

Both prompts test the same core skeleton (1 entity, 4 attributes, filter by category, 1 derived value, page refresh state, single user) with a deliberate ambiguity.