AI System IDS Achieves Formal Verification of Distributed Systems

A new AI system, Inductive Deductive Synthesis (IDS), has successfully generated formally verified distributed systems, a task where previous state-of-the-art coding agents failed. IDS jointly and incrementally synthesizes implementation and proof, learning from failed attempts to systematically explore promising strategies. Built as an agentic LLM system, IDS achieved 7/7 on distributed key-value-store specifications in about 6.8 hours at a cost of $106. In contrast, SOTA coding agents Codex with GPT-5.4 and Claude Code with Opus 4.6 succeeded on only 2/7 specifications. The work addresses a critical gap in AI's ability to provide formal guarantees of full coverage, which testing alone cannot ensure, particularly for properties like consistency under all possible event interleavings. Mechanized formal verification typically requires months to years of expert effort. The paper is available on arXiv under the title "Inductive Deductive Synthesis: Enabling AI to Generate Formally Verified Systems."