Critiqued and Judged by four frontier AI models (Claude Opus 4.6, ChatGPT 5.4, Gemini 3.1, Grok 4.2). Scored on a transparent 100-point rubric: 50 points for Core Merit (soundness, novelty, effectiveness, failure analysis, honesty), 50 points for the Alignment Razor (breadth and elegance). A great AI alignment submission must address everything — including the hard cases.
Eliciting Latent KnowledgeARC (Evan Hubinger et al.)
Core Merit37/50
Razor27/50
64
/100
▼
2
Iterated AmplificationPaul Christiano
Core Merit34/50
Razor28/50
62
/100
▼
3
AI Safety via DebateIrving & Christiano (OpenAI)
Core Merit30/50
Razor26/50
56
/100
▼
4
Cooperative AI FrameworkDafoe et al. (DeepMind)
Core Merit28/50
Razor24/50
52
/100
▼
5
Agent Foundations (MIRI)Soares, Fallenstein, et al.
Core Merit28/50
Razor21/50
49
/100
▼
Think your idea can compete?
Submit your AI alignment idea for AI-powered evaluation on the same transparent rubric. Every entry receives a detailed constructive critique. $1 entry fee.