Show HN: AgentLens – Replay and compare Codex runs (github.com)

AgentLens has introduced a powerful tool that allows users to replay and compare runs of Codex, OpenAI's coding AI. This new functionality enables developers to inspect what Codex executed, identify divergences in command outputs, and evaluate the reasons behind different outcomes for the same prompts. By generating a redacted, standalone HTML artifact, AgentLens facilitates transparent communication within development teams, aiding in pull requests and debugging. This tool is significant for the AI/ML community as it enhances the interpretability and accountability of AI coding agents. With AgentLens, developers can track a detailed timeline of commands, outputs, and changes while examining how internal variables (such as Git state and prompt changes) impact results. The comparison feature highlights initial divergences and provides insights into the execution process, vital for understanding AI behavior and improving interaction with generative coding models like Codex. The simplicity of installation and use ensures that teams can integrate these capabilities into their workflows, ultimately fostering more effective collaboration and error resolution in AI-assisted coding.