Did my Custom GPT just show signs of AGI? /joke

The title is not clickbait, just a joke. But…

I have a Custom GPT connecting to an API I wrote. The API has add, delete and update item functionality, pretty typical. I was testing just the add/delete endpoints from my GPT and had only added those to the instructions/schema.

Add and delete worked as expected, but I forgot that I hadn’t told it about update yet, and I tried to do an update. It found a workaround.

It first deleted the item, then added it back with the updated values - two endpoint calls. Has anyone else seen that kind of behavior?


That’s a nice experience.
I recall having a long conversation about some C module and when adding a new function a data type was changed.
The model proposed the change and added that a previously written function would have to be updated as well to accommodate the changes.
I was actually happy that I didn’t have to use another prompt for this adjustment. But yes, I consider this to be somewhat a sign of intelligence.

Ps. I just wish this would happen more often :rofl: