AI coding assistants like Claude and Cursor have let junior developers pick up complex languages like TypeScript in weeks instead of months. But code review is still slow, and adoption is wildly uneven: some engineers get ten-fold velocity gains while others barely use the tools. The bottleneck has shifted.
Chapters
I mean, I've seen like even in our company, like a lot of the developers, like especially junior developers, they don't like code like the way I used to code like 5 years back. Like now they're like really going to ask Claude Code a question, even a simple question to explain the code base. They're not navigating the code anymore. They're just always defaulting to go to AI. The barrier has been lowered. Like now what we're seeing is the earlier the barrier was like you have to learn a new language, like you don't know TypeScript. If let's say your company is using TypeScript, you have to first learn that, or Rust or something, right? Now what we're seeing is like developers are now very comfortable, like even picking up Rust, for instance, which is very hard otherwise, right? They're letting AI do a lot of the heavy lifting, and through AI-generated code, they're also like picking up those skills faster. In a way, like we've seen acceleration in a lot of areas, like especially the junior developers. I've seen like a lot of these like younger 20-year-olds have been very, very successful with AI compared to people who have been coding for like 20 years. So for them, it's like a bigger change. Newer habits. Whereas like by default, all the new grads we hire and all, they're like much more like more accustomed to using these agents, right? And they're also finding like innovative ways these agents can be used. Like, I think one of the things is that we have seen even the model labs don't even know the full extent of where these models are more successful or less successful. Like, if you look at this coding, like, I mean, a lot of people are now like doing this plan mode, like cursor added that, like just to be more successful or writing markdown files with task lists. So everyone's building their interesting workflows around these agents to be successful, right? And these agents are now running for longer and longer periods of time. And also describing that destination where you want to be is an art in itself. I've got— at CodeRed, we have so many developers, and I look at their usage stats in Cursor or Cloud Code. You could actually see how much they're doing on-demand usage. You will see there are a few developers which are really going gangbusters with AI usage while And some of the developers are still struggling and they're doing basic tab completion and so on. So it's very interesting to see, it's very lopsided. Like, it's very hard to even tell what makes, like, what kind of a trait you will look for a developer that they'll be good, like, really leverage these AI tools to the maximum extent. Because that makes a 10x difference in the velocity, right? I mean, if you're gonna use these new tools, you can obviously run faster.
