It should be embarrassing to people that they can’t even write a whole argument out by hand. It’s not that hard. All you have to do is believe in it.
This blog post is just selling the idea of engineering as craft. It exists to make us feel good about having skills and to get us to regurgitate our priors in the comments. We should at least demand that the sales pitch be manual.
So, I immediately write-off anyone who produces their writing with LLMs, as it is quite apparent that they are not a serious, thoughtful person and are instead just cosplaying as one via LLM. The fact that they can't see this (or, worse, think the slop actually makes them look good) only further proves the point. I'd much rather read flawed but genuine writing over vacuous AI slop, and surely most people are the same.
The other day someone posted in reddit about a repo that they seem to have put a lot of thought into. But the post and all comments were AI slop. I told them that while the project looked promising, I couldn't take them seriously on account of the slop writing.
They replied thanking me. Then the next day followed up to say that I changed their life - no more Ai-generated writing when interacting with other humans. I hope they were serious about it.
I’ve recently started writing again, after well over a decade, so I’m not very up to date on how things are done and perceived. FWIW, I write these on Google Docs and then port them over to my site. Earlier posts had an AI disclaimer saying I used Google Docs which presumably uses AI for spellchecking and obvious grammar issues but it ticked people off so I’ve stopped adding those.
The harness though is real; the LLM does not do anything interesting without a set of tools, specific prompts to use them, and taking a lot of forced context planning out of the dev's hands.
But yeah, usually when you mix business interests without customer concerns, those business interests develop dark patterns so they can invent concerns and then solve them.
It's still hard to stop yourself from constant bikesheding, reorganizing, etc, but yeah, mentally it's a different bag of tricks you need to learn.
Usually, I start with a solid idea that needs hands on experience to fully developin, so I need to stay in the loop. Than I have to force myself to carefully review the code and point out architectural issues (usually there sneak in a lot of them). If I don't force myself, I end up in a limbo where the agent produces code faster than my mental model following.
Even if I have a well developed domain knowledge and a good idea what I want, I then definitely feel some kind of, what I guess compares to, slot machine effect. "Maybe now I get what I want. Argh, not quite. Okay, feels nearly there. Perhaps only this prompt. Ah, there is something missing...".
I'm really curious how you handle and approach this.
It got me back into enjoying things more
