For reference (as per Wikipedia):
Any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the organization's communication structure.
— Melvin E. Conway
Imagine interpreting that as advice on how you should try to design things, lol.
Tbf, I think most of the post is just typical LinkedIn fluff, but I didn't want to take the poor fellow out of context.
That version is so much more readable for some reason. Like this is actually a wonderful sentence:
That doesn't sound (to me) like an LLM, and yet the stuff it produces by default reads in a tone that we would have described as "robotic" long before GPT.
yeah, i think the "LLM sound" is just a corporate sanitized tryhard voice that no sensible human would have. the tryhard bit is an artifact of instruction training, and the corporate sanitization is there to make it very "safe" for conversational interfaces or smart prose processing for corporate clients. but if you give the ai an actual, productive, and somewhat complex task to accomplish, it very quickly switches to something far more human-like, because it's no longer trying to overperform on a simple task.