Recently stumbled on the post “Friction Was the Feature” by Ethan Johnstone and it really stuck with me. Worth reading here.
Roughly his point is writing something good used to take real effort, and that effort had unique value to people “consuming” the writing. A well-written application or pitch meant somebody cared enough to put in the time or had the “skill”. Now that AI can produce “pretty good” almost instantly or with enough iterations, that signal is gone. Everything looks fine (sort of). It’s harder to tell who’s actually good anymore based on the general quality of what they created.
It first reminded me of pitch decks honestly. A genuinely high quality presentation used to be a decent way to filter out the consultants and businesses who didn’t know what they were doing, mostly because faking that level of polish took almost as much work as actually being good. Big consultant companies probably re-used these pitch decks (often forgetting to change client names) because of how they “signaled” depth and readiness to the client.
Really the same thing applies to code. Writing clean, well tested, maintainable code was hard enough that being able to do it consistently gave you an indication of skill and confidence in a code base. Junk code that only worked on the happy path of execution was a signal of a problem.
That’s not as evident anymore. AI can write code that looks clean and passes the tests without anyone involved fully understanding why it works, including the person who technically wrote it. The quality bar didn’t go away, just stopped implying the same thing.
I guess the old expression of “buyer beware” is more true than ever when we can’t rely on the same filters to evaluate the value of things.