Why We Run 9 LLMs in Parallel Instead of One (And Sign Every Output with Post-Quantum Crypto)

Short summary

ENLIL runs up to 9 LLMs in parallel isolation, then synthesizes their independent responses into a single signed output called a Decree. Unlike sequential pipelines, models never see each other's outputs during deliberation, eliminating anchoring bias. A review round allows anonymous cross-critique before final synthesis. Every Decree is signed with ML-DSA-87 post-quantum signatures for tamper-proof auditability, targeting high-stakes use cases like security reviews and legal analysis.

• •9 LLMs deliberate in parallel isolation, synthesizing independent responses into a signed Decree
• •Models never see each other during deliberation; optional review round adds anonymous cross-critique
• •Outputs signed with ML-DSA-87 post-quantum crypto for tamper-proof audit trails

Generated with AI, which can make mistakes.