March 31, 2026
Popcorn, not patch notes
Good Code Will Still Win
Good Code Will Still Win? Crowd splits between 'ad' cries and 'AI will solve it'
TLDR: Greptile argues clean, simple code will win because it’s cheaper to build and maintain, even as AI helps people ship more. Commenters erupt: cynics say good code never won, futurists claim machines will “solve” software, others call the post an ad—while mods ask everyone to play nice.
“Good code will still win,” says a Greptile post arguing that simple, easy-to-change code will beat messy AI-generated “slop” because it’s cheaper to build and keep running. They point to a world where developers ship way more lines thanks to AI helpers, outages are creeping up, and even tech celebs like Ryan Dahl say the “era of humans writing code is over.” The pitch: markets hate mess, and clean code saves money.
The comments? A full-on food fight. One camp is deeply cynical: “Good code never won before, why would it win now?” They say the industry’s always been “ship fast, don’t touch it later,” and AI won’t change that. Another camp goes full futurist, insisting machines will soon have massive memory and so much feedback that software becomes a “solved problem”—and yes, more complex can be better, like a super-fast jet. Meanwhile, a pragmatic voice shrugs: if “good” means “useful and cheap,” then sure, that wins—think the cheapest bridge that doesn’t fall down.
There’s also drama over tone: one commenter calls the post “basically an ad”, and a moderator jumps in with a “be nice” reminder and links the guidelines. The memes write themselves: it’s jets vs. bridges, slop vs. savings, and hope vs. habit. The only thing everyone agrees on? More AI means more code—and whether that code is clean or chaos could decide who ships and who sinks.
Key Points
- •Greptile argues economic incentives will drive AI models to produce simpler, maintainable “good code” because it is cheaper to generate and maintain.
- •Greptile’s 2025 State of AI Coding report indicates lines of code per developer rose from 4,450 to 7,839 as AI tools became standard.
- •Median PR size reportedly increased 33% from March to November 2025 (57 to 76 lines changed), with individual files 20% larger and denser.
- •The article cites an analysis suggesting outages have steadily increased since 2022, implying growing software brittleness amid rising code volume.
- •Citing Ousterhout’s design principles, it contends complex code is costly in tokens and compute, while simpler code reduces context and modification costs.