LLM as Interface

I’m old enough to remember when graphical user interfaces appeared on desktop machines in the 1980s. Many of them were wrappers around pre-existing programs; while some programmers (including my younger self) sneered at them for being pointless eye candy, all of a sudden ordinary people didn’t have to remember a long list of obscure command-line spells in order to make computers do useful things.

I’m now old enough to remember when LLMs appeared in the 2020s. All of a sudden people didn’t have to learn how to write syntactically correct YAML or figure out which of Git’s eleventy jillion flags would safely and correctly do whatever they wanted to do. I’m now wondering whether “LLMs make it feasible for non-specialists who aren’t heavily invested to use X” is the same as “X’s interface never actually needed to be so user-hostile”, and whether “talk through it with an LLM” is going to make complex software more accessible in the same way that GUIs did.