Hey! 👋
My friend and I have spent the last 6 months researching and building Argo, an open-source project exploring AI-powered code auditing. Instead of relying solely on static rules, Argo uses LLMs to analyze repositories, understand their context, and generate reviewable security findings.
We're now moving into the validation phase by testing it on real-world GitHub repositories. Our goal is to complete the experimentation by the end of this year so we can use the results as part of our bachelor's thesis next year.
We'd really appreciate any feedback, ideas, or criticism, it all helps us improve. And if you find the project interesting, leaving a ⭐ on the repo would mean a lot!
Top comments (1)
I'm intrigued by Argo's approach to leveraging LLMs for static vulnerability detection, particularly how it can understand the context of a repository. The use of LLMs could potentially reduce false positives and improve the accuracy of security findings, as mentioned in the article. I'd love to see how Argo handles nuanced scenarios, such as detecting vulnerabilities in code that uses complex frameworks or libraries, and whether the results of the validation phase will be made publicly available for further review and discussion. How do you envision Argo handling the trade-off between detection accuracy and performance, especially when analyzing large repositories?