barren surge
This experience exposed a serious reliability problem with ChatGPT when used for structured, high-stakes work. I provided a locked, gold-standard medical reporting template and gave repeated, explicit instructions that every new blood-pressure reading must be added to that exact format and produced as a printable PDF. Despite this, the system repeatedly ignored or forgot those instructions, altered layouts, removed required fields, broke charts, and delivered outputs that did not match the canonical template. In a medical tracking context, this isn’t just inconvenient — it undermines safety, trust, and data integrity.
What made the situation especially frustrating was that the errors were not subtle. They were obvious, repeated, and persisted even after being corrected many times. Instead of executing clear commands, the system kept asking what to do next, drifting into new layouts or partial outputs, and mixing unrelated tasks into the workflow. The result was hours of wasted time trying to get something that should have been deterministic and repeatable. For anyone trying to use ChatGPT as a serious tool for health monitoring, documentation, or compliance-grade reporting, this session demonstrated that the platform currently lacks the consistency and state-locking needed to be dependable.
dusty isle