A public company CEO told me AI coding has had negligible impact on his engineering teams, instead the real transformation has been on their product and design teams using Replit. I asked him how does he reconcile this with CEOs saying that 25-50% of code is generated by AI? He said that’s also true in their case—AI does generate a lot of their code—but that whatever time saved in generating the code is lost back in debugging, reverting bugs, and security audits. So if you measure time to ship, PRs merged, or whatever high-level metric you don’t see any impact. Whereas his non-technical teams gained a fundamentally new super power of being able to make software. Prototyping with Replit makes iteration speed incredibly faster before it gets to engineering. And non-product teams—like HR—can for the first time solve problems where vendors don’t have the exact solutions they’re looking for. I was surprised to hear the part about engineering teams, and I’m sure every company will be different, but it made sense the profound impact coding agents are having on non-technical folks.