Buzz Andersen @andersen.buzz · Feb 14 Not sure I agree with this characterization. Another way of looking at it, as this essay explains, is that AI takes over the creative, generative part and essentially turns the human job into code review. Which seems like what it’s going to do to everything. siddhantkhare.com/writing/ai-f...
59 likes 5 replies
Replies
Yeah, I think for folks who use it to generate slop code that they then foist on others, that’s correct. (And bosses are counting on that.) But the most recent wave of tools are changing that possibility for the other end of the spectrum of folks who are anti-slop.
The tool is what you make of it. I've had apps on the shelf for years because, after a work day, I don't really want to write a bunch of data management and plumbing (and tests for it). Those projects are getting new life because of LLM coding, and I can focus on the fun and interesting UI problems.
The reviewer/judge framing is right. What makes it worse is the evaluation never resolves cleanly. Not just "is this code correct?" but "is this output worth refining, replacing, or scrapping?" That's a decision every single iteration. Stacks up fast.