The terrifying liminal space between code-writer and bot-prompter

October 17, 2026

Disclaimer: These are my own thoughts and opinions. They don’t reflect those of my employer. I’m discussing my own personal feelings about AI. I think it’s fair to say everyone is having to grapple with this tech, whether it’s how to use it or how to avoid it. The following represents a snapshot of my grappling, nothing more.

LLM coding advancements both impress and terrify me and I don’t know what I think about it or how to cope with it. I feel like I will be able to build things I never have been able to before, but at the same time, I have heightened fears about being replaceable or less valuable in my job. If all I’m doing is prompting an AI bot, why am I being paid so much?

When I work with an LLM coding harness^[1] it usually makes me feel uneasy. It feels like my name is going to be attached to something I didn’t really do. Sure, I could review the code. But it’s still not my code. Constantly reviewing LLM-generated code is mentally fatiguing and is not enjoyable in the way coding is. And at a certain point, it really becomes a challenge to keep up with code review. Before long, you find yourself skimming and saying, “That seems right.”

A developer who is skilled with and/or enjoys working with this stuff would probably say you don’t need to review the code. Trust the LLM, trust the tests. As much as I hate it, I think that might actually be right. Trying to keep up pace with a bot that can spew out thousands of lines of code in minutes seems like a fool’s errand.

The models are good now. That’s the refrain I’ve been hearing for the last six months. Again, as much as I hate to admit it, I have to agree the models are much better than they were. I’m routinely impressed, particularly in the GPT models. A year ago, hallucinations seemed to be more of a problem. Nowadays, the LLM will search the internet, search the code, and run its own experiments before taking a stab at answering your questions. I don’t know if it’s the models that are that much better, or if it’s a huge improvement of the harnesses. Probably both.

You’re probably sick of hearing this, but if you haven’t used one of these coding tools in the last six months, you probably don’t have a good idea of what they are capable of.

But I hate it. Why do I hate it? I just have this icky feeling when I use it. And that’s not even considering all the other big negatives like how it kinda feels like these companies stole from everyone. Or the negative environmental effects. Or all the mindless slop that’s everywhere now.

There’s obviously a lot of doom-mongering in the industry. The companies making this stuff say AI is going to take our jobs or, I shit you not, kill everyone. So that’s… good.

But let’s assume for a moment that the worst-case scenarios aren’t true. What does software development look like in the future? I’m worried about being replaced, but maybe I’ll be even more valuable. If the cost to produce software drops, Jevons paradox says people will consume even more software. If demand for software goes up, well, I’m someone who can produce software (or begrudgingly facilitate a machine producing it). I have soft skills and insights around my technical knowledge, and maybe that continues to be valuable.

So I just go back and forth. I worry about it, but then I think everything will work out. It’s something that’s been on my mind a lot this past year. I can’t just quit tech to become a gardener. As someone with spinal muscular atrophy, some kind of knowledge work is all I can do. The government programs that I use for my daily care don’t cover 100% of my needs. Without my income, I would be very near needing to live in a nursing home.

So I’m keeping watch. I’m staying open minded and trying to find a path forward in this new world.