It’s incredibly ironic that since AI is trained on buggy code it’s very good at writing buggy code.
-
It’s incredibly ironic that since AI is trained on buggy code it’s very good at writing buggy code.
Truly a monkey’s paw situation for AI productivity gains as you are now producing bugs at a faster rate.
-
It’s incredibly ironic that since AI is trained on buggy code it’s very good at writing buggy code.
Truly a monkey’s paw situation for AI productivity gains as you are now producing bugs at a faster rate.
@carnage4life wait, I'm lost: you tend to post a lot of positive messages about coding agents, but that one here gives the feeling that you consider those to be crappy tools... what am I missing / what part am I misunderstanding?
-
It’s incredibly ironic that since AI is trained on buggy code it’s very good at writing buggy code.
Truly a monkey’s paw situation for AI productivity gains as you are now producing bugs at a faster rate.
@carnage4life think about how bad average code is and realize that half the code is worse than that
-
@carnage4life wait, I'm lost: you tend to post a lot of positive messages about coding agents, but that one here gives the feeling that you consider those to be crappy tools... what am I missing / what part am I misunderstanding?
@gturri People on Mastodon are used to every debate being a religious debate and so someone discussing the pros and cons of any particular technology is confusing and causes cognitive dissonance.
Coding agents are a tool, like a web browser or virtual machine. They make a bunch of things easier but then introduce unique problems as well.
-
It’s incredibly ironic that since AI is trained on buggy code it’s very good at writing buggy code.
Truly a monkey’s paw situation for AI productivity gains as you are now producing bugs at a faster rate.
I think this is more fundamental problem than most folks.
If you are generating 2x more bugs, you will not catch 2x more bugs you will catch maybe half of the additional bugs.I am actually going to assume the rate is lower on teams that heavily rely on the tools b/c they are likely not doing deep code review and will likely catch a lot less.
Long-term this will become a much larger drag on project velocity than any gains you could make. I don't think this is controversial here, all of these basic concept we live by don't go away just b/c we have fancy tools.
There is no free lunch
. For those prototyping, you are the exception b/c you will throw away the code you do indeed get a free lunch. That is a rather small exception though. -
It’s incredibly ironic that since AI is trained on buggy code it’s very good at writing buggy code.
Truly a monkey’s paw situation for AI productivity gains as you are now producing bugs at a faster rate.
@carnage4life true, but think of all those out of date-yet-up voted stack overflow answers from a decade ago blindly copied and pasted even though there were later answers pointing out how flawed the original was
At least with SO you can see the corrections -
E em0nm4stodon@infosec.exchange shared this topic
