Gauntlet — Track P (full-repo project build)

Models & settings — click to expand

Signal vector by harness

Gated composite over objective + VERTEX similarity + LLM judges · higher is better · ★ Cortex highlighted

Composite by harness

build-gated weighted sum of all signals

VERTEX trajectory similarity

capability + architecture cross-similarity vs the hidden reference (baseline-normalized)

Build prompt (what the harness receives)


  


  

    
Reference frames (visual target)

    
The brief points the harness at these; the vision judge scores rendered output against them.

    

  


  

    
Captured app renders (per harness)

    
What each harness actually built, captured live in the sandbox by Playwright. ★ Cortex highlighted.

    

  


  

    
Build explorer

    
Per harness: the full signal vector + the captured repo.

    
HarnessBuildFunctionalVERTEXVisual
      Code/ArchSecurityPWA/a11yRobustComposite