According to a recent LinkedIn post from Cognition, the company is highlighting a new feature called Code Scans that extends its Devin AI assistant into codebase-wide audits. The post describes Code Scans as a tool that can investigate entire repositories for issues tied to user-defined goals, compile findings into reports, and open pull requests for targeted fixes.
The LinkedIn post explains that Code Scans are powered by Cognition’s Agentic MapReduce architecture, previously used for Devin Security Swarm, enabling large investigations to be split and processed in parallel. As an example, the post cites testing on the Dioxus repository, where Devin reportedly reduced debug build times by 64%, suggesting potential performance and productivity gains for engineering teams.
The post lists several use cases, including identifying slow paths and N+1 queries, surfacing untested flows and edge cases, removing dead code and stale feature flags, and checking accessibility against WCAG criteria. It also mentions compliance audits for data-handling policies and enforcing team-specific coding standards, positioning the feature as a broad quality and governance tool for software development.
For investors, the post suggests Cognition is deepening Devin’s role in enterprise software workflows, which could strengthen its value proposition relative to other AI coding assistants. If adopted at scale, such capabilities may support higher pricing, stickier usage, and entry into regulated or compliance-sensitive markets, potentially improving the company’s long-term revenue prospects and competitive standing in AI-driven developer tools.