Your AI Coding Assistant is Lying to You - Here's the Proof
I used to think Copilot made me faster. Most people I work with say the same thing. Then I read a study that ruined my week, and a few more after that, and now I am not sure anymore.
This is what I found. I am sharing it because I think a lot of us are wrong about this, and the data is not hard to look up.
A study that should make you uncomfortable.
In July 2025, a lab called METR did something simple. They got 16 experienced developers. Real ones, with years of work in big open-source repos. Some of these repos had over a million lines of code. The kind of codebase you can't fit into any context window.
They gave each developer 246 real tasks from their own repos. Bugs to fix, features to add, normal stuff. For each task, a coin flip decided if the...
Copyright of this story solely belongs to hackernoon.com. To see the full text click HERE
