Hacker News new | ask | show | jobs
by birdsongs 23 days ago
> Don’t take my word for it then, ask any terminal agent to dig in and get an idea of how good these apps are.

Okay but, you know this isn't a quality metric right? These models are incredibly biased towards positive confirmation of the prompt. I could give them nearly any repo and they would sing the praises of the best parts, if I asked.

This sounds a bit like psychosis.
1 comments
theturtletalks 23 days ago
I never told you what prompt to even feed your AI my guy. As far as I know, you can be like rip this repo to shreds and tell me what’s wrong with it and you’ll get your answer.

Crazy how you guys blindly trust proprietary apps where you can’t even read the code but asking you to read open-source code is psychosis?

link
birdsongs 23 days ago
> Crazy how you guys blindly trust proprietary apps where you can’t even read the code

Wow okay, I'm a huge open source advocate, I run hardly anything closed source. No one here was talking about that, this is your own weird segue.

> but asking you to read open-source code is psychosis

But you didn't ask that. You told me to get an agent to read it (actually, not even read it, make its own quality assessment for me).

That's a massive difference, and if you consider them to be the same thing, then yes, my statement stands.

link
theturtletalks 22 days ago
You say you only use open-source and I’m the same way but I’ve asked agents to evaluate open-source alternatives. How is that psychosis? If it’s typescript or JavaScript, I can read the code but if it’s Rust/Go/etc, yeah I get AI to read it and tell me if it can solve my problem or if it has the features I need.

It’s actually the whole premise of my open source alternative directory called Opensource Builders[0].

0. https://opensource.builders

link