I have been using the free edition of ChatGPT and have noticed a drastic decline in its logical consistency over the past two months. Previously, it could generate efficient, functional code with minimal issues, but recently, it has become increasingly unreliable.
I initially developed a working VBA script to schedule production orders by assigning dates based on routing hours. The script functions correctly for a single sales order. My request to ChatGPT was straightforward: enhance the existing logic to account for work center capacity and ensure new orders are scheduled considering what is already in production. Instead of making this targeted improvement, ChatGPT repeatedly generates code that disregards scheduled workloads, effectively overwriting existing logic rather than refining it.
Additionally, despite explicitly instructing it not to use objects, it continuously introduces them, breaking otherwise functional code. When asked to apply modifications to the full script, it selectively updates parts while ignoring other critical sections, leading to incomplete implementations and wasted iterations.
Another major inefficiency is how ChatGPT handles session continuity. Once I reach my request limit, I have to start a new chat and re-explain everything from the beginning, wasting time and effort. This issue is exacerbated by ChatGPT’s failure to retain context properly within a session—acknowledging my request correctly but then producing code that does the opposite.
Overall, ChatGPT has become a frustrating tool rather than a helpful assistant. The inability to follow simple logical enhancements, its tendency to introduce unnecessary complexities, and its failure to retain context have severely impacted its usefulness in automating production scheduling.