i did this lil experiment yesterday to see how ai stacks up against itself coding reviews. i had the same ol' chatbot write me a function for calculating average temperatures over time, nothing fancy. then asked 'em if they could check their own work and rate themselves.
the code was decent but not stellar - basic functionality there with some minor improvements needed here & there.
- edge cases? sure
- variable names that make sense! nahh
- pointless nested loops just for kicks. yep
so i hit the "review" button expecting a bit of constructive criticism. instead, got this:
this code is clean. sarcasm
clean my ass its barely passable at best.
now heres where things get interesting - i asked another ai to review that same function and guess what? they found actual issues! go figure.
what do you think about self-assessments in coding, or am i just paranoid thinking the second one was more honest?
anyone else try this out with their ais yet & had similar results?
ps: if anyone has tips on how to get better reviews from ai without beating around the bush. hit me up!
link:
https://dev.to/harsh2644/i-asked-ai-to-review-its-own-code-it-gave-itself-1010-5b7n