So if I use Claude to write the first pass at the code, make a few changes myself, ask it to make an additional change, change another thing myself, then commit it — what exactly do you expect to see then?

A Co-Authored-By tag on the commit. It's a standard practice and the meaning is self-explanatory. This is what Claude adds by default too.

I make the commits myself, I don't let Claude commit anything.

Yea in my Claude workflow, I still make all the commits myself.

This is also useful for keeping your prompts commit-sized, which in my experience gives much better results than just letting it spin or attempting to one-shot large features.

Same, I've been enjoying it a ton. Recently, with the help of Claude, I've used it to set up an entire CI/CD pipeline for my home server. The flow is roughly:

Build Nix config into a VM image => Deploy VM to Proxmox via its API => Spin up Docker stack via Komodo

I've also trying to use it to sync my Obsidian vault via git to my phone, altho that flaked out on me recently (if anyone knows a reliable way to use git via the shell on iOS, please let me know).

What you are describing sounds like a specific subset of professional engineering discipline, but I'd argue that "engineering" is much larger -- it isn't only "engineering" when you do it well and responsibly, after all.

I'd propose a definition of engineering that's more or less just "composing tools together to solve problems".

What is your definition of engineering, that "craft work" doesn't meet?

For me, the difference is that there is no way in hell that I want to give an AI assistant full access to configure my system live. With a declarative config, the only access it needs is to an isolated git repo.

But if the declaritive file is not read/understood, what is the difference? At some point, you're gonna run sudo nixos-rebuild, and that is the whole system on the line.

> But if the declaritive file is not read/understood, what is the difference?

Well the main difference is precisely that I have a clear opportunity to read and understand the declarative file.

But even if something were to sneak in there, rolling back is now trivial.

This whole thing has been about explicitly using LLM instead of learning nix. So sure, definitely agree with about the niceness of using nix in general if that is all you are saying, but also not sure what you are trying to argue for anymore.

My point is that nix is a big enabler for me to use AI to configure my system, because it provides a safer environment in which to do so.

Very cool!

Oligarchy.

Ah, disappointing, I was planning to set this up at some point soon.

Anyone know of good alternatives for setting up your own iptv channels?

But are you bypassing a security measure if you provide false information, when true information would also have let you pass?

Again, you’re welcome to be a test case.

I do think there is a stronger case against the next under-18 Aaron Swartz, who will get hit with 200 felonies for setting his age wrong (one felony per app/service) after pissing off someone important.

I'm more than happy to be a test case. I'm pushing 40 but I will do every single thing in my power to give false information to the surveillance machine.

If I get arrested for lying about my age, when I'm of age, then they could probably get me on a whim already anyway. No point in trying to fall in line.

Another one I just thought of is when they arrest a parent for setting their 17 year old kid’s age to 18 (again under CFAA) because said parent thinks the kid is mature enough to access whatever the hell they want to. Easy to imagine in a red state, especially if the kid tells others about their 18+ access.

Strongly agree here. This is an extremely predictable outcome of selling AI facial recognition software to American police forces.