Trying to build a CHAT GPT prompt which can be easily modified to analyze wide variety of student loan situation. Same prompt does not always give same results, Very troubling. Requesting your input.
Published a paper with CHAT GPT yesterday. Created a prompt that allowed me to replicate all tables. Verified that Python and Excel calculations were IDENTICAL and correct. Paper is here.
Reran prompt today and got worksheet error messages, different results in CVS files and gibberish in Summary files.
here is a picture of the verification/summary table from yesterday.
Here is a picture of the summary verification table from today.
Same exact prompt, different answers, and various explanations from CHAT on what was happening. here’s one.
Totally. Since your Excel setup behaves better with named ranges, I built you a V4 that uses names everywhere in the Summary (no explicit cell/area references). That should be the most robust for your environment.
Am trying to create a prompt that can be modified to address several different types of student loans. Need Excel versions to verify Python calculations but sometimes Excel models built with help from CHAT don’t work Well.
Your thought? Have you had similar experiences?
I had similar issues when trying to compare student loan payments and taxes owed in 2026 as married filing jointly or married filing separately. different outcomes each prompt!