Untitled

I like how Nolan paints this two-sided picture of AI coding. The worst fact about these tools is that they work. They can write code better than you or I can, and if you don’t believe me, wait six months. You could abstain out of moral principle. And that’s fine, especially if you’re at the tail end of your career.

https://chriscoyier.net/2026/03/02/13264/

@chriscoyier "can" seems to be mentally transposed with "will" by executives, and it elides the fact that code generation getting cheaper means that the new most important skill is knowing to ask for *good* code (for whatever version of "good" matters in your domain). These tools spit out crap by default, and that won't change because the corpuses they're trained on are pretty average.

A sense of taste and occasion, combined with understanding what's possible, is the new (old) differentiator.

@slightlyoff @chriscoyier "This is so much cheaper and faster" works for employers up until the AI suppliers jack the prices up. Which they will, and they'll go up by "as much as is most profitable for us, the AI supplier, which the market will bear". And the market bore salaries. Guess where the price point will head into?

And yeah, we still need to know code. I will tell you *as an end user* of programs: there are certain ones I *need* solid human oversight of the code. For trust reasons.

@slightlyoff @chriscoyier yup 100% this. The biggest danger to engineering careers right now isn't AI, it's business leaders who think AI can eliminate engineering.

The difference between unguided random production and the output from an experienced engineer with an LLM is stark. But what we're going to see in the short to mid term is many exec leaderships pushing to eliminate engineering orgs and that'll work kinda if average or below average output is sufficient which for many it might be?

@bendelarre @slightlyoff @chriscoyier This is what I might call Revenge of Enterprise-y.

~20 years ago, a lot of “enterprise software" was simply ghastly, exhibiting terrible UX and an absolute pain both to write and to run. Much leaner-and-meaner frameworks & languages & practices emerged among hackers/startups (Ruby on Rails, for example). Mobile “wow" arrived on the scene. Enterprise-y simply wouldn't cut it. UX won…

…until now.

Slopware is the new Enterprise-y. UX & process now bad again.

@bendelarre @slightlyoff @chriscoyier While "vibe engineering" causes all roads to lead to a stinky middle of Next/Tailwind soup, poor affordances, brutalist design, and uncertain outcomes, teams laser-focused on handcrafting *excellence* will run circles around increasingly dysfunctional FOMO-fueled orgs who've lost the ability to gauge what good UX even is.

Oh yes, I am *directly* connecting the dots between LLM-avoidance & UX, because they're two sides of the same coin.

And I have receipts.

@bendelarre @slightlyoff @chriscoyier Well said. We need engineers more than ever if the result of all that code generation is meant to be reliable useful products. Of course that's not the way it's going to go. But the people who see this coming have an opportunity to differentiate themselves in the long run.
@slightlyoff "taste and occasion" is good. almost as good of a blog name as "infrequently noted"