Comment by eru

8 months ago

The Claudes are a lot worse at even mildly challenging algorithmic problems than Gemini 2.5 Pro.

However, most legacy code is fairly primitive on that level, so my observation is in no way contradicting yours.