AI and Coding: It’s Not What You Think
The real win isn’t writing more code — it’s working with the code we already have.
The media is selling a flashy story about AI churning out endless lines of code like some tireless junior developer. But as Laura Tacho reminds us in her excellent interview with The Pragmatic Engineer ( Measuring the impact of AI on software engineering), the real revolution isn’t in code generation — it’s in code comprehension.
The full interview is an hour and 11 minutes long and covers a range of topics/perspectives that I’d broadly characterize as debunking AI myths in software engineering. Great inputs!
At Bad Science Fiction, I’ve been working quietly with my AI consultants toward a relaunch at some point in a new direction. (Call experiment 1.0 mostly wrapped.). And from these perspectives, if I had to pull one decisive insight from Laura’s interview, it’s this: code generation isn’t the big win of AI. The real value lies in breaking the bottlenecks of code comprehension — debugging, analysis, understanding (including its intersection with design).
As Laura notes at about the 27-minute mark, typing speed has never been the real bottleneck in development. Churning out more code simply shifts the effort to reviewing, interpreting, and understanding it. AI’s biggest win won’t be in generating endless code—it will be in helping us work smarter with the code we already have.
The subtext to Laura’s discussion is that there is a lot of misinformation circulating in the media about AI and its impact on coding/development. Similarly, on the macro-political side, Noah Smith has an excellent opinion piece up, pushing back against AI pessimism (again, fueled by misinformation). To move past the hype, we need to focus on where and what AI truly delivers value—bridging the understanding gap.