Desarcasmed version: "I think that problems with Zork make those models virtually useless in programming tasks." Correct?
replyHe said complicated code bases. LLMs are great at producing small snippets of code to address very targeted problems.
replyGreat on small snippets of code, passable on larger pieces of code, great at finding vulnerabilities in large pieces of code, terrible in Zork. All-in-all, a jagged frontier that defies a simple sarcastic characterization.
replyVery kiki, not very bouba, as Aphyr rightfully stated.
reply