6/15/26
People like to say “AI is a Force Multiplier”. This is the popular concession that shows up when I say my coworkers are pumping out a whole lot more bad code with this new tool. The implication in this is that our good engineers are also able to produce a lot of good code with the tool.
I’m skeptical of this claim. I think that this tool’s primary use case is to produce lots of bad code. I still do not really believe that it can produce lots of good code for a good engineer. I haven’t really seen proof that it can generate lots of good code in any way.
You can try to bully it into creating good code, by repeatedly re-prompting it. Or try to edit its bad code into good code. I am skeptical that this process is actually an efficient use of time or energy.
The only argument that I can see making sense is “we shouldn’t look at the code any longer”. If all code should just be bad code, then maybe there’s a real story here. Thus far, I have not seen this to be the case. In my work, I always end up needing to understand the bad code when it breaks or needs tweaks and then I wish that it had just been good code from the outset.
Maybe we’ll get to a place where this is no longer true. I do not think we’re there yet. I don’t want to hear about “how it will work next year”. I want to know how to deal with the dumpster fire I’m surrounded with right now.