• 0 Posts
  • 26 Comments
Joined 9 months ago
cake
Cake day: June 11th, 2025

help-circle



  • LLMs are shit a doing large code changes, and fundamentally will always be shit at it, because they fundamentally can’t reason. LLMs are good text completers, and that’s their place in the IDE.

    My prediction is that most well run organizations are going to push against coding agents soon. Look at the reports that even Amazon is now demanding that Senior Engineers review for AI code changes and take responsibility, which doesn’t scale now, and will scale even less in the future if we train less coders.