A Survey of Bugs in AI-Generated Code

This is an academic survey paper (0 stars, 5 forks, 123 days old, zero velocity) that performs empirical analysis and categorization of bugs in AI-generated code. It is not a software tool, framework, or deployable component—it is a literature/empirical review that documents findings about code quality issues. DEFENSIBILITY: Score of 2 reflects that this is a survey/research paper with no practical software artifact, no adoption pathway, no users, and no defensible IP. The contribution is observational and taxonomical, not a novel tool or technique. Replicating the methodology requires only applying existing quality assessment techniques to AI-generated code—not novel engineering. PLATFORM DOMINATION: Low risk. This is not a platform capability that would be absorbed into AWS, GCP, or Azure. It is academic analysis. Code-quality tooling (linting, static analysis) is commoditized, and platforms already have DevSecOps/code scanning offerings. A survey of bugs does not threaten platform roadmaps. MARKET CONSOLIDATION: Low risk. No incumbent vendor is competing to own "bug surveys in AI-generated code." This is domain research, not a commercial product category. Code quality tools (SonarQube, GitHub Advanced Security) are mature and separate from this analysis. DISPLACEMENT HORIZON: 3+ years (effectively unlikely for this artifact). Surveys document findings at a point in time; they do not compete with tools or services. However, the broader domain of AI code-quality assessment will evolve as models improve, making any specific bug taxonomy increasingly outdated. TECH STACK & INTEGRATION: This is a research paper with an accompanying reference implementation. It is not pip-installable, not an API, not a CLI tool. It describes methodologies and findings that could be implemented by others, but provides no composable software artifact. COMPOSABILITY: Theoretical. The value is conceptual: a taxonomy of bug categories and recommendations for evaluation. Researchers could implement their own tools based on the taxonomy, but the paper itself is not a reusable component. IMPLEMENTATION DEPTH: Survey. No production system is being described. Code examples and analysis exist, but the contribution is observational. NOVELTY: Novel combination. Applying empirical analysis and bug taxonomy to AI-generated code is a meaningful synthesis of established software-quality evaluation methods applied to a newly relevant problem domain (AI code generation). The novelty is in framing and scope, not in the underlying techniques. CONCLUSION: This is academic research with limited defensibility or commercial threat exposure. It serves as a reference for researchers and tool developers but does not itself constitute a product, platform, or defensible asset. Its value is in informing future work on AI code quality—not in direct adoption or competition.