More AI-written code is not better code
AI speeds up development, but the challenge now is understanding and maintaining what gets written.
Artificial intelligence is already speeding up software development. According to a new study by the Linux Foundation (LF Research), AI can raise productivity by up to 50%, and developers estimate that writing code takes up to 52% less time. But the cost shows up elsewhere: 59% of open source project maintainers report growing contribution volume, and an estimated one in four pull requests (25%) is low quality, what the report calls "AI slop".
The challenge is no longer producing code faster, but verifying it, governing it, understanding it, maintaining it and being able to trust it. Code that nobody understands is a debt that gets paid later, in every fix and every change. That is why 46% of maintainers say AI increases the need for mentoring, versus 32% who say the opposite, and only 35% of projects have a formal policy on AI use. For us, the conclusion is clear: review, testing, documentation and clear rules for AI-assisted code matter as much as the code itself, because what is not understood cannot be maintained. How is your team handling it?