On 01/06/2026 11:03, Kieren MacMillan wrote:
Hi all,
there is no.guarantwe that the AI agents actually reflect appropriate
understanding of the code base.
Adding them to LilyPond will add cognitive debt, which I believe is much
worse than technical depth. When AI -generated code is created, there is a
strong likelihood that nobody in the world understands why that particular
code works and is an appropriate solution for the problem under consideration.
A valid and important point. To paraphrase the brilliant Cory Doctorow: “Code
is a liability — not an asset (as most people seem to believe) — and AI lets us
generate that liability at scale.”
Butting into the discussion late, but if we want a simple question "how
much of this code was written by AI", then I would think the following
is a fairly simple way (if contributors are honest) of giving us a good
idea of the code quality and whether it's worth reviewing:
" Did you use AI to help in the creation of this work?
If so, was the primary use of AI
(1) assisting in debugging code that you wrote (including code from)
(2) getting AI to write the first draft before rewriting it yourself
from scratch, or
(3) writing the bulk of the submitted code, including debugging that code "
Both (1) and (2) give assurance that the contributor has a good
understanding of the code they are submitting. I've made (a little) use
of (1), and I've had somebody else give me (3) which I turned into (2).
Both experiences were positive.
But (3) should ring alarm bells. That code I got from my fellow employee
was completely unfit to commit "as is". You can then push back on that,
and tell contributors that they need to rewrite it, show that they
understand it, and submit it as a patch series rather than one big patch.
The problem, of course, is that people who submit a big (3) patch are
also probably the people who are least inclined to take any request to
break the code up and justify it with any good grace :-(
Cheers,
Wol
