Funny enough, I see it the exact opposite. LLM slop code degrades software quality. While one person may appear to move quickly, the software quality suffers dramatically. It takes them (or other team members) more time to fix the issues than it would have to just do it right the first time.

I mean, even with the advent of “AI” agents, is software today any better than it was 3 years ago?

I think teams embracing too much LLM coding are going to see a continued decrease in quality and a massive drop in team productivity. Godot, on the other hand, is avoiding this. While it might not be as trendy, I think it’ll be a competitive advantage in the long term.

I don't know about large teams or companies. I feel more confident with the conjecture: Conditioned on small teams, a team of seniors with AI coding beats a team of junior with AI coding by a wide margin, in my experience.

When accounting for "coding taste," I think there is equal or higher quality output over small teams, and if there is a decrease in software quality, it's nowhere near the implied amount (5-10% at most) with a massive gain in speed (approx 3x? 4x? hard to measure). Those numbers net out positive in the end, especially as the team starts to understand and ship with AI in the loop.