Ensuring that software behaves exactly as intended is a constant challenge, especially in complex systems. Traditionally, proving code correctness involves intricate manual proofs, a process akin to a logic puzzle on steroids. But what if AI could shoulder some of this burden? Researchers are exploring precisely that with "AutoVerus," a new tool that leverages the power of large language models (LLMs) to automatically generate correctness proofs for Rust code. Rust, known for its focus on safety and performance, is increasingly popular for systems programming. Verus, a verification tool designed for Rust, allows developers to write specifications and proofs directly within their Rust code. AutoVerus builds upon this, using LLMs to generate the complex proof annotations needed to ensure code aligns with its intended behavior. AutoVerus works in three phases, mimicking how human experts construct proofs. It begins by generating an initial, preliminary proof. Then, using common proofwriting strategies, the tool refines this initial attempt, adding details and correcting oversights. Finally, in a debugging phase, AutoVerus addresses any remaining errors, drawing on a collection of specialized LLM agents each trained to handle specific types of verification issues. Imagine an LLM agent meticulously examining the code, comparing it to the formal specifications, and adding the necessary logical steps to bridge the gap, much like a seasoned programmer anticipates and addresses potential bugs. One of the key challenges is the scarcity of training data for this kind of task. Unlike more established verification tools, Verus has a relatively smaller pool of examples for LLMs to learn from. To overcome this, AutoVerus relies on a network of highly adaptable LLM agents, each specializing in a different aspect of proof generation or error correction. The results are impressive. Evaluated on a benchmark suite of 150 proof tasks, AutoVerus can automatically generate correct proofs for over 90% of them, often in mere seconds. This opens exciting possibilities for more efficient and reliable software development, particularly in areas where code correctness is paramount. AutoVerus is not just generating proofs—it's paving the way for a future where AI-powered tools help us build more trustworthy and secure software systems, automating the tedious yet essential task of ensuring our code does exactly what we expect.