Switched from GPT-4 to Claude for coding help and it's night and day

Been using AI to help debug my Python scripts for about 8 months now. Last week I tried Claude for the first time on a problem with pandas dataframe merging that I'd been stuck on for 2 hours in GPT-4. Claude walked me through the merge logic step by step without me having to rephrase my question 5 times. Has anyone else noticed big differences between these models for actual coding work?

3 comments

3 Comments

derek93919d ago

Wait, so you're telling me I've been asking GPT-4 the SAME question five different ways like some kind of digital parrot and Claude just... gets it the first time? That's honestly hilarious, especially since GPT-4 keeps asking me "are you sure you want to merge on this column" like it's my first day coding. I swear, sometimes I feel like GPT-4 is just trying to sound smart while Claude actually wants to solve the problem. The difference is like asking a fancy chef to boil water versus asking your buddy who actually cooks. Claude doesn't make me feel dumb for not knowing the exact magic words to get a straight answer.

sean_foster5219d ago

Yeah I've definitely been that guy asking five different ways before too.

johns1814d ago

Wait, did your buddy ever try asking GPT-4 to just "fix the CSV merge" and get some rambling answer about data normalization? My friend Mike was banging his head against the desk last month trying to get GPT-4 to help him join two sales reports in Python. After 45 minutes of rewording his question like 8 times, he just quit and played video games. Next day he tried Claude and it was like a different universe - Claude saw the mismatched column names and actually showed him how to rename them first. He finished the whole script in 15 minutes and texted me a screenshot of the working output with like 20 exclamation points.