Hi everyone, I'm trying to help my team pick a primary AI coding assistant. We're a team of five engineers focused on building and maintaining Python ETL pipelines, mostly using pandas, PySpark, and SQLAlchemy. We deal with a lot of messy data transformations and scheduling.
I've read all the feature lists and marketing pages for both Claude Code and Windsurf, but honestly, I'm more confused now. They both say they're great for Python and understand context. The pricing models are different, which matters for our budget, but I care more about day-to-day usefulness.
Can someone who's actually used both for similar work share a concrete example? Like, what happens when you give it a real, gnarly task? For instance:
* You have a legacy script that uses an old version of a library and you need to upgrade it and handle the deprecated functions.
* You ask it to write a robust error-handling wrapper for a series of API calls to a third-party service.
* You need to refactor a slow pandas `apply` function into something vectorized.
I want to know which one "gets" the problem faster, gives fewer hallucinated library methods, and produces code that actually runs with less back-and-forth. How do they handle the boring but critical stuff, like adding proper logging or writing a clear docstring?
Our stack is pretty standard, but the devil is in the details. Any real-world anecdotes would be super helpful to cut through the hype.
Help me decide