“What is so confusing about using LLMs for coding right now is that you feel like you can solve any problem… but you don't end up with this really neat core design that has the least amount of possible states. It looks like it does, but it actually doesn't.”
Fragmented - AI Developer Podcast · Hands-On Engineering Podcasts · October 2026
Armin Ronacher on why designing a small durable-execution library with Mario Zechner took repeated rewrites. Zechner had just said that you hit roadblocks with every design.
Transcript
us really long to feel good about this.
And it's not like it's amazing in the sense of something entirely new. It's just you have to think through so much stuff that you're hitting roadblocks with every design. 100% eventually. And I think
what is so confusing about using LLMs for coding right now is that you feel like you can solve any problem and you can sort of do, but you don't end up with this really neat sort of core design that has the least amount of possible states if you like, it looks like it does, but this actually doesn't.
This is one of the most tempting things. That's why I started with like, do you guys have some kind of philosophy? Because in this day and age, it's so tempting, right? Like it's so tempting to just throw an agent at it to solve the problem versus stepping back and thinking, no, no, but is this the…
I mean, again, it comes back to this. Can I keep it in my head? And I can keep PyDurable in my head for the most part. Obviously, some details, like how does the bash tool concretely handle, I don't know, Windows versus Unix-like systems in terms of process groups and child trust. I don't remember…
You briefly touched on like, you know, process ownership. And that's one of those questions I had as well. Because funny enough, you know, Mario, I know you released your sub-agent extension. I've had my own sub-agent Py plugin that's been there for some time. If you think about it from the surface…