Chatbot programming: An example

  I wrote,

I have a list of 4-tuples of str objects.
These 4-tuples are called "rules".
A rule is a full-word rule if its component 0
starts with "<" and its component 2 ends with ">".
The complexity of a rule is the length of the str object
that results from concatenating the first three components
of that rule.

Write code to sort this is list as follows: First the
full-word rules ordered by decreasing complexity, then
all the other rules ordered by decreasing complexity.

  , the chatbot replied:

|def is_full_word_rule(rule):
|    return rule[0].startswith("<") and rule[2].endswith(">")
|
|def rule_complexity(rule):
|    return len(rule[0] + rule[1] + rule[2])
|
|def custom_sort(rules):
|    # Separate full-word and other rules
|    full_word_rules = [r for r in rules if is_full_word_rule(r)]
|    other_rules = [r for r in rules if not is_full_word_rule(r)]
|   
|    # Sort each list by decreasing complexity
|    full_word_rules.sort(key=rule_complexity, reverse=True)
|    other_rules.sort(key=rule_complexity, reverse=True)
|   
|    # Concatenate the sorted lists
|    return full_word_rules + other_rules

  . The code of the chatbot is better than the code I would have
  written manually, because it's sweet how he represented my English
  concepts "full-word rule" and "complexity" by a def each, which I
  would not have the patience to write myself had I written the code
  manually! I'd just have inlined the expression in "custom_sort".
  I also might not have written such nice comments . . .

  Note how it helped to make the code and comments meaningful that
  I created a mini terminology in my prompt.

  (Disclaimer: Bryce Adelstein Lelbach [and maybe Frances Buontempo and
  many others too] already used ChatGPT in 2023 to generate code.)