OpenAI has described a system it calls the Performance Factory, a set of workflows powered by Codex, its AI coding agent, that are designed to find and fix performance problems affecting ChatGPT. The episode of OpenAI's podcast covers how the system is built and why the company treats responsiveness as an ongoing engineering task as usage grows. The approach applies automated code analysis to the routine work of keeping a large consumer product fast. Rather than relying only on engineers to notice slowdowns, the workflows are meant to surface issues and propose fixes that can be reviewed and shipped. The show notes do not include detailed benchmark figures, so the size of any gains and the specific bottlenecks addressed are not clear from the available material. For Australian businesses and professionals, this changes nothing directly. ChatGPT's performance is not a setting they control, and the development does not bring new costs, contract terms or Privacy Act obligations. The indirect relevance is that faster, more reliable hosted AI tools make it more practical to rely on them day to day, so response times are worth checking when comparing vendors, but no action is needed now.