🔗 Reviewing AI code is not a viable argument

A strong skepticism towards LLM Coding Assistants arises from their fundamental flaws in code generation. Thomas Depierre critically examines the argument that thorough code reviews can mitigate the issues with AI-generated code. He highlights that relying on these tools often feels like mentoring an inexperienced intern. The inherent limitations of LLMs, including inaccuracies and inconsistent performance, lead to excessive human review that may not yield the efficiency promised by proponents. Empirical evidence underscores that a senior developer can only effectively review a limited amount of code in a fixed time frame, severely hampering productivity. As such, the suggestion to simply review all AI-produced code raises more questions about its practicality and effectiveness than it answers. The call for empirical research on the efficiency of human reviewers handling LLM output remains critical for validating such tools' efficacy in software development.