AI council finds critical bugs missed by top models in smart contract audit

A developer created a local multi-agent AI system called Egregor to audit smart contracts, finding it significantly more effective than using individual AI models. Egregor, which uses a council of five AIs including Gemini, DeepSeek, ChatGPT, Grok, and Claude, identified four critical bugs in a smart contract that multiple individual models missed over thirteen manual iterations. This structured approach, costing approximately $0.40 per run via the Openrouter API, utilizes role distribution and an "Anti-Groupthink" engine to improve accuracy and efficiency in code review. AI

IMPACT This approach could significantly improve the reliability and efficiency of smart contract audits, reducing risks for Web3 projects.

RANK_REASON The item describes a new tool for AI-assisted code review, not a release from a frontier AI lab.

Read on dev.to — LLM tag →

• ChatGPT
• Claude
• Claude Opus
• DeepSeek
• Egregor
• Gemini
• Gemini Pro
• Grok
• Openrouter API
• SovereignBank Web3

• product
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →