AI Written Code Not That Clean

Coders are committing more incorrect code, then coming back to correct it later

GitHub CEO Thomas Dohmke in June 2023 posts a blog with the extraordinary claim that CoPilot would add $1.5 trillion to GDP:

Too good to be true?

Of course, statistics like these are often calculated the same way lost productivity due to the Super Bowl is calculated, without thought to the balancing game theory that counters the effect. And sure enough, the good people at Gitclear evaluate these claims in detail:

Code Churn in light blue

Gitclear identified code “churn“:

• percentage of code that was pushed to the repo, then subsequently reverted, removed, or updated within 2 weeks by the same author

• normal churn in 2020/2021 was about 3.4%

• increased to a projected 7.1% in 2024

In effect coders are committing more incorrect code, then coming back to correct it later.

Gitclear draws the conclusion that this is AI-driven behavior, which I agree with, even though the evidence here is circumstantial at best.