Ran into some limitations with Chat GPT 3.5 on the weekend. I was trying to get it to take a list of X numbers that represented the sat amounts of a group of UTXO's, and get Chat GPT to:

1) count the number of figures in the list provided, as a reference point, to make sure none were missed

2) group the UTXO's into separate groups whose sums were all as close to a predetermined amount, making sure not to reuse or eliminate any of the figures from the originally supplied list of figures.

The program started by not being able to count the number of figures in the list correctly. I had to correct it, after which time it agreed that I was correct and it had miscounted them. Then followed two hours of tweaking instructions to try to get it group the UTXO's into roughly equal larger amounts, close to value Y.

It failed miserably. It was outputting groups whose values were way above and way below the target sum. And when asked to refine the grouping, it was still leaving some groups way off the target value.

I was able to go through the list and manually do some reasonable, relatively equal groupings in 15 minutes, but I thought it would have been cool if it had been able to just crunch it for me.

A project for another day, using dedicated language and functions I guess. It must be a pretty simple piece of code.

One question that arose during the excercise was if I alternately grouped the highest value figures with the lowest value UTXO amounts, reducing the number of smaller UTXO's that appear in a given group to be consolidated with a single transaction, would that have positively impacted my fees for the whole operation? I am thinking yes from a data size perspective, but I lack a critical piece of knowledge about how much data is used for each UTXO. Anyway, it was a relatively low fee day so I didn't get raped too hard.