Hacker News Reader

Comment by poszlem | Hacker News Reader

"Mostly Claude Opus 5.5 in Claude Code, with one human steering by prompts. "

I find it funny that I would happily use a version of this if I created it (with Claude), but somehow cannot get myself to use it since it's been prompted by someone else. I wonder what's that about.

View on HN

That's something I don't fully understand with a lot of these (clearly ai generated) projects.

If you didn't take the time to build the thing you're trying to sell to me (even if it's free) why should I buy (use) it? I could just prompt my own solution instead of randomly curl-installing some thing from someone I don't know.

Am I the crazy one here?

Well, tokens cost money, and in theory, if someone else already built something you need that costs a billion tokens, it makes more financial sense to use theirs than prompt your own.

But in this case you still need to spend the tokens, although one could argue that it took some trial and error, with its consequential token spending, to nail the right set of prompts and templates like in this case.

> I could just prompt my own solution

Okay but in the limit, you can't prompt all the things. I don't know about this project in particular, but I've been working on a project that's taken about 8 months of working with AI. Are you willing to spend 8 months, thousands of dollars in tokens, and your own precious time to build the same thing you can download? How many projects are you willing to do that for? Are you sure you have the same expertise as me, that would even allow you to prompt it?

It reminds me of when everyone was trying to build their own frontend framework. Like "neat, I agree with this in concept, but I don't trust your code or you to maintain this project"

I feel the same way, there’s a disconnect between writing the code and prompting the code that requires a different form of interaction from the human, I’ll call it respect for lack of a better word.

The problem for me, and I suspect for you, is I don’t know how to quantify the level of respect other humans have put into their prompt engineering and iterative processes to ensure the output is satisfactory.

I think we’re going to eventually learn how to determine prompt competence the same way we’ve determined code competence but it’s going to take a while before it all shakes out.

I think that's it really, the fact that someone else made it. Sometimes we make things for ourselves that lean perhaps on the _functional_ side, (tools, food...) that we would never want served to us. And AI just exacerbates this. I do wonder why that is though.

Very true with the food analogy. I’m not eating whatever this persons concoction is, but I’ll make it myself at 3am and chew it gloomily.

For me it's because I've seen what others do with Claude vs how I iterate with Claude. There's an enormous range in the bound "one human steering". Oh, is that a zen koan? What is the sound of one human steering?

Reply on news.ycombinator.com