Besides that, I don't think that the "appeal to authority" the author makes in his article works. Wirth was a pragmatist, not a luddite. His "Lean Software" philosophy was about determinism and understandability, i.e. building systems where the engineer comprehends every instruction cycle. He criticized software that was written poorly (from his very own perspective).
The author implies that because we ignored Wirth’s warnings about OS bloat in the 90s, we were culturally primed to accept the "ultimate bloat" of LLMs in the 2020s; that's pretty wild logic. The author essentially criticizes that using LLMs for simple tasks is overkill. This is another concern than writing bloated software.
LLMs have demonstrated that they are able to generate both, lean and bloated software. When I use an LLM to review my code, it often finds logic reductions and possibilities for minimization which were not obvious. Ultimately, it once again comes down to the question of how tools are used. In the case of LLMs, they can obviously be used in line with or contrary to Wirt's recommendations.
I'm not particularly sure what the autor's claim has to do with Wirth, but there is a quote which he could have added to the article: "To hell with AI" (see https://inf.ethz.ch/people/faculty/emeriti/niklaus-wirth-quo...). Wirth distrusted AI mostly because of the "grandiose promises" its proponents made at the time (see https://bertrandmeyer.com/2024/01/16/niklaus-wirth-importanc...).
Besides that, I don't think that the "appeal to authority" the author makes in his article works. Wirth was a pragmatist, not a luddite. His "Lean Software" philosophy was about determinism and understandability, i.e. building systems where the engineer comprehends every instruction cycle. He criticized software that was written poorly (from his very own perspective).
The author implies that because we ignored Wirth’s warnings about OS bloat in the 90s, we were culturally primed to accept the "ultimate bloat" of LLMs in the 2020s; that's pretty wild logic. The author essentially criticizes that using LLMs for simple tasks is overkill. This is another concern than writing bloated software.
LLMs have demonstrated that they are able to generate both, lean and bloated software. When I use an LLM to review my code, it often finds logic reductions and possibilities for minimization which were not obvious. Ultimately, it once again comes down to the question of how tools are used. In the case of LLMs, they can obviously be used in line with or contrary to Wirt's recommendations.