Post #4112145

2026-07-26 12:47 UTC

@levitte@mastodon.nu: "What I have a harder time is the argumentation that's used, and quite specifically that agentic coding leads to slop almost by definition. " Almost? What _is_ the definition os "slop"? There is no objective measurment of code quality, but I'd say "legacy code" is just code that no human cares about enough to understand. By that definition, llm-generated code is already legacy code when produced, and that's a good enough definition of "slop" for me.

Replies (3)

@rkaj@mastodon.nu 2026-07-26 12:48

@levitte@mastodon.nu Some people may nog agree that "slop" is always bad. But saying that all llm-generated code is slop is consistent with any definition os "slop" I use.

Open ##4112155
@ahltorp@mastodon.nu 2026-07-26 13:12

@rkaj@mastodon.nu @levitte@mastodon.nu I would add that all generated code, whether by LLMs or compilers is immediately legacy code. That is why it's so important that generated code can be easily regenerated at any time. This includes machine code. Editing generated code has always been a bad practice, and it doesn't suddenly become a good practice just because we have LLMs. For exploratory programming, the trade-offs are a bit different, but by definition maintainability is not an issue in that case.

Open ##4112571
@levitte@mastodon.nu 2026-07-26 13:32

@rkaj@mastodon.nu https://codeberg.org/ethical-foss/open-slopware uses a self-fulfilling definition: > slop | Anything generated by AI, typically code, books, articles, or images. It's pretty dumb and nonsensical if you ask me.

Open ##4112905