The Speed of Writing Code Is Probably Not the Problem ⇥ andrewmurphy.io
The speed of writing code was never your problem. If you thought it was, the gap between that belief and reality is where all your actual problems live. The competitive advantage doesn’t go to the team that writes code fastest. It goes to the team that figured out what to build, built it, and got it into users’ hands while everyone else was still drowning in a review queue full of AI-generated PRs that nobody has the time or the energy to read.
Via Elizabeth Ayer:
The fact that we are *not* seeing wildly improving software all around us tells us everything we need to know.
There is no flourishing of value delivery, new product categories, more needs being satisfied better. It’s the opposite.
All we are seeing is decreases in quality, because 👏 code 👏 creation 👏 is not 👏 the problem.
Nilay Patel, making a tangentially related point on Bluesky:
I keep saying “there are no great consumer AI products” and people keep replying to me with like model capability updates and wild OpenClaw setups and I really fear Software Brain is irreversible
The iPhone was a consumer product so great that enterprises were forced to adopt it! That’s the bar, not the other way around.
I completely agree with Murphy’s argument from a professional perspective. Though I write limited code these days, I want to understand it by developing it myself. The bottleneck there is a quality-based one. I need to know what I am building, and what bugs I have created so that I may create something better. I cannot get that through generated code because, as for anything automatic, I will stop being attentive.
But for personal projects, the bottleneck is absolutely a function of available time. Little side projects sit there until I have ample time to solve them. For example, MarsEdit has a lovely little bookmarklet that will start a new post containing the highlighted text. For years, I had been meaning to modify it to Markdown-encode any emphasized text and set links in my preferred reference style. My JavaScript skills being quite rusty, I knew that was going to require ample time that I did not want to spend. So last year, I threw it at ChatGPT, and it did an admirable job of updating it to my needs.
I am conflicted about this. I decided to avoid learning something and judge the output solely based on whether it works as expected. And, to Patel’s point, I felt like I was using a corporate tool for some hobbyist project, which is unpleasant. It has solved a point of friction in my workflow — not itself a bottleneck, per se, just something I found a little bit annoying.