Except, LLM-gen'd and human-written code are different.
One was written by someone with a mental model. Modified by humans with another mental model (correct or not), etc.
The other is generated by statistical, stochastic methods with constraints.
These are not the same.
Want to use LLM-generated code? That's your choice. But don't argue that it's the same as code from 15 years ago, written by folks who had a human mental model, and made sense to them at the time.
@jitterted it’s easiest to see if you have an unorthodox solution to a problem. Then LLM would try hard to refactor your code to the more popular solution, or worse, assume the popular solution without noticing the actual one, leading to massive refactors or even logical bugs. Because it does not have mental model, it just „sees” something that differs from the usual (under some definition of usual).
Jun 21, 2026, 8:39 PM@jitterted Low-background code https://en.wikipedia.org/wiki/Low-background_steel
Jun 21, 2026, 9:04 PMHuman-written code is written with the intention of the code working. Machine-generated code is generated without intention. The machine does not care if the code works.
It's difficult to imagine how this would not result in two wildly different sets of code, even if both share the same output. And one of those sets of code is likely to be intrinsically much more possible to maintain.
Oooh, I'm adding a screenshot of this reply to my talk. Love it!
@d_stepanovic Right?? Intent, judgment, even emotion enter into our coding activities, which an LLM will never imbue.
@jitterted
100%
...and I'll trust agentic coding (to some extent at least) as soon as OpenAI and Anthropic accept liability for consequences of the code their models generate.
Pretty much the same expectation as I would have from Tesla and their claims of 'fully autonomous driving'.
I.e., likely never.
Jun 22, 2026, 12:29 PMBy all means! Though made an adjustment -- 'possible' to maintain, not 'likely' to maintain. I was obviously quite tired on the first pass!
Thank you.
Quoted here as well
https://www.linkedin.com/feed/update/urn:li:share:7474864307407802368/