You could say the same about patents for physical inventions. But there are legitimate reasons for patents to exist.

Nice strawman.

It’s a term AI bros came up with to describe replacing the entire development lifecycle with AI.

Software factories

🤮

the code you have to review hasn’t been thought about

Isn’t that just admitting that the author of the PR has not read the code they’re submitting? If I caught wind of that I would reject the PR on principle without even reading all of it.

If you are competent and using an LLM to generate code you’ll put your name to, that mean reviewing a lot of code that isn’t up to your usual quality, rather than just writing it yourself, which is frankly extreme tedious, demoralising, and mentally exhausting

That’s not been my experience. While there are often multiple iterations of feedback I go through with an LLM before I consider a PR ready, it’s rarely a significant deviation from what I already planned out. Most often it’s just basic code organization that I forgot to specify (especially if the area of code it’s working on was already poorly organized), or tradeoffs I wasn’t even aware of until the code was written. The amount of effort per cycle is usually very small, and I’m able to work on other things in between (like reviewing other peoples’ code).

If you’re not competent, you wont catch any of that, and it’ll either just go to production, or it’ll fall to someone else to review it.

This sounds mostly like a workplace culture problem. It’s always been important to have a high bar on code review to catch poor quality. Not to catch bugs per se, but to catch code that isn’t properly tested.

you’re going to end up flooded with low quality commits from coworkers who’ve either given up or weren’t competent in the first place

Again, that’s a workplace culture problem. You need to either stick to your principles and try to enact change, quit, or go along to get along. I’m sorry if the last option is where you’re at and that does suck.

it’s much harder to train people to good practice when they’re abdicating the thinking to a machine

I totally agree. That’s something each engineering org needs to reckon with, lest they burn out the devs who care about code quality.

Spending your day reviewing code churned out by an LLM, is a soulless and tedious way to spend your life.

Someone has always had to review the code. And the author has always been responsible for their PRs meeting a quality bar. I don’t see how this is any more soulless or tedious than before we had LLMs. Maybe you just never enjoyed reviewing code in the first place, and now you have to do it more?

I think you’re thinking about this in a very negative way when you also have the option to see it as an opportunity. You said yourself that now you’re focused more on design and architecture, and most of your time is spent reviewing code. That sounds like a relatively good place to be, assuming you’re happy with the quality of the code. Now you can spend more time solving new, harder problems that you couldn’t before.

Do y’all remember StumbleUpon?

As I witnessed today, even just stating that you are opting out is not enough. You must ensure, of your own accord, that the camera is not directed at you when you approach the TSA agent, or else it will automatically photograph you.

I’ve also had a TSA agent tell me that I need to tell them I opt out before handing them my ID. This is obviously problematic but that’s just how this shit works today.

Wow small world. Have you played Spitzer?

I wonder though if they just told agents to literally reverse engineer Photoshop. As in, they were given access to Photoshop and able to interact with the UI and figure out every minute behavior. With a good testing strategy, I could see agents producing a decent quality product. That’s basically strip mining years of effort put in by humans.