News for hackers. Stuff that matters.

anfal

anfal on Hacker News

Re:Ask HN: Anyone frustrated by AI tools hallucinating API calls? on Tuesday April 08, 2025 @03:04PM (#43625386)
Yeah that's exactly what I am talking about. Like I was working on this livekit project and I asked cursor, could you help me override this method? it very confidently gave me the incorrect answer and instead of speeding things up, I had to spend 2 hours just to figure out how do it by digging in their docs.

I am just curious here, would this be something you'd be willing to beta test? while I build it? I am dogfooding my own system and I wanna work with 5-10 people who have felt this pain and want to do something about it.

Re:Ask HN: Anyone frustrated by AI tools hallucinating API calls? on Tuesday April 08, 2025 @02:24PM (#43624920)
Just to clarify — I'm not trying to promote a product here. I'm genuinely trying to understand if others are running into this problem too.

I’ve talked to a few devs who said their AI tools confidently suggest incorrect SDK usage, especially when docs change frequently.

They also mentioned that they either manually copy and paste the relevant doc page (highly unoptimized) or keep the docs opened in parallel in a browser tab.

Would love to hear if that's happening to others, and if so, how you're working around it. Manual doc digging? Custom tools?

Also happy to go into how I’m approaching the solution if anyone’s interested.

Ask HN: Anyone frustrated by AI tools hallucinating API calls?

Posted by anfal on Tuesday April 08, 2025 @02:23PM
from the ask-hn dept.

I’ve been using ChatGPT, Claude, and other LLM-based tools to speed up software development — but I keep running into the same issue:When integrating third-party APIs or SDKs, the AI often gives me outdated or completely wrong info. Sometimes it’s minor (wrong param), sometimes it burns hours.I figured the root problem is docs go stale, but the AI still tries to be confident. So I’m building a...

( Read More... | 3 comments | on HN )

Show HN: Making SDK/API docs usable in LLMs (zero setup for companies)

Posted by anfal on Monday March 24, 2025 @07:26PM
from the show-hn dept.

I'm building ChatVisible AI — a tool that makes dev-focused companies (offering SDKs/APIs) instantly LLM-ready.It turns documentation into LLM-usable, up-to-date knowledge — so developers can use AI tools (like Cursor or Copilot) to build faster without hitting outdated or missing docs.It’s LLM-agnostic, requires zero setup on the company’s side, and gives devs a much smoother experience when...

( Read More... | no comments | 2 points | on HN )
Re:Is Hallucination a Vehicle for Creativity? on Wednesday February 19, 2025 @01:54PM (#43105861)
I've been exploring whether AI hallucinations are simply failures or if they might resemble human creativity. The article looks at how hallucinations in AI could serve as a generative mechanism, similar to brainstorming or lateral thinking. Also touches on reinforcement learning as a way to guide hallucinations into productive outputs. Would love to hear HN’s thoughts?
Re:TL;DR of Deep Dive into LLMs Like ChatGPT by Andrej Karpathy on Sunday February 16, 2025 @06:32PM (#43072943)
Fully agreed, someone actually mentioned that to me on reddit and I modified the content + added a disclaimer on top of the article.

https://www.reddit.com/r/LocalLLaMA/comments/1ilsfb1/comment...

Re:TL;DR of Deep Dive into LLMs Like ChatGPT by Andrej Karpathy on Sunday February 16, 2025 @06:29PM (#43072920)
Hey, Anfal here. I am actually the author of this article. I have had some really good discussion with a few really intelligent fellows I know and they tie very deeply into deepseek more. I'll create a post about it soon.