Discussion
Some months ago, I saw a post here in Mastodon about a developer who had quit AI in their job. I found it very inspiring. Here is mine.
Not without shame, I admit that for a long time I used AI at work, even though I had banned it from my personal life. Now, what I believe and what I do are aligned again.
If you're a developer, please read it. There is life after being an AI sheperd.
https://blog.bustikiller.com/2026/09/25/one-month-without-ai.html
@bustikiller I'm very proud of you. I do think the addiction metaphor is applicable to LLMs, and that it's a mark of personal strength to overcome it.
The other metaphor I've been considering lately is that traditional development is a bit like reading a book, and agentic development is like watching a movie (or with multiple agents, several movies at a time on different screens). If you have a lapse in attention while reading the book, you can find your place again and re-read. Maybe you need to set it aside for awhile and nothing gets done, but nothing gets "not-read" either. But movies, they keep playing when you tune out, and when you tune back in, you're lost.
I have projects that have gone very slow, or had long pauses, because I'm disabled. An LLM could have given me the power to auto-pilot through those lapses in ability, but I'd never be able to catch up and clean up afterwards.
@MaddieM4 thanks for your kind words 🤗
I like your metaphor as well. In my opinion, important things take time, have moments of growth and moments of pause. The same way summer cannot last forever, flowers don't bloom all year, and projects (and people) have their ups and downs. And that's fine, that's good.
The obsession for productivity that AI companies are trying to force us into goes in the opposite direction, and IMO hurts the project (and the people) in the long term.
> When you take a draft PR written by AI and have to work on it yourself, you realize to what degree you didn’t know what it was doing.
This matches my experience and it's been the most striking thing to me. It's not just that the AI got things strictly wrong, but _weird_, the bikeshed is all the wrong colors. And it's super hard to notice until you start reworking with the code by hand
@bustikiller Really interesting read, thanks.
As someone who's not a programmer, but a very experienced sysadmin (who obviously does *some* programming!), the idea of relinquishing control of my systems to an unknowable, unseeable, unaccountable "thing" is anathema - as you came to realise.
@bustikiller Thank you for sharing your experience with this, it was a very interesting read.
Install bonfire.cafe
Get the full app experience
@bustikiller
This is encouraging.
I recently left my job, partly due to ethics, but also due to AI mandates.
Even if I opt out of the tools myself, I have to deal with code-reviewing peers who did not. Reviewing LLM code IMO is trickier. And for that extra effort, it's less rewarding than the growing together experience of reviewing human code.
I've been thinking I'll need to leave the field entirely, but if more people have this realization, maybe I might be able to stay.