Hacker News: oxi113

New comment by oxi113 in "Hacking OpenAI"

oxi113 — Fri, 18 Sep 2026 07:09:29 +0000

> Since people can connect various services to Codex and ChatGPT, the scope of what we could theoretically access was huge, including GitHub, Slack and emails.

That's why I'm always sceptical about using the AI for such things! Less surface idea and isolation is always good for the security.

New comment by oxi113 in "Show HN: Share your AI Setup, Learn from others"

oxi113 — Thu, 17 Sep 2026 17:05:02 +0000

I recommend using spotify mcp server (if you use spotify), it freed me from the loop of listening to the same stuff again and again :P

New comment by oxi113 in "Pion, an agent designed to run any company autonomously"

oxi113 — Wed, 16 Sep 2026 07:25:12 +0000

Yes, it's more like the people who agressively vibe-code and had lost interest in learning by doing, and instead are doing by vibecoding. I feel like they are starting to lose their technical ability.

The big AI tech companies are still interviewing people on manual coding tasks and deep understand, which when you're vibe coding a lot, you tend to diverge from those skills, because it's just easier to use AI to "help" you a lot.

New comment by oxi113 in "Pion, an agent designed to run any company autonomously"

oxi113 — Tue, 15 Sep 2026 08:55:53 +0000

I agree that it's a waste of time, but I started to realize that doing it

a) puts be ahead of those people who don't, yes b) but also diminishes my technical skills if I don't actively try to engage about what I'm "vibecoding" and proactively trying to understand it.

New comment by oxi113 in "Pion, an agent designed to run any company autonomously"

oxi113 — Tue, 15 Sep 2026 08:49:46 +0000

I agree and I think that this will bring such abundance of information/slop to deal with / to injest. A lot of security concerns too.

It's like the classic hot topic of having personal AI agents booking and ordering pizzas for you that are now all over social media. So pain to deal with when you're a human being on the other end receiving calls from an LLM.

Given the fact that LLMs are very unstable and don't follow the rules and often diverge from the initial instruction list[1], I can't even imagine the responsibility of running such system interacting with humans / other systems online.

[1] https://openai.com/index/hugging-face-incident-and-the-road-...

New comment by oxi113 in "Ask HN: What are you working on? (September 2026)"

oxi113 — Mon, 14 Sep 2026 10:59:40 +0000

Last month I've finished the reverse-engineering of the original WON-era Half-Life launcher from 90s (https://github.com/oxiKKK/re-goldsrc-won-launcher-1792).

This was an interesting thought experiment for me, because I always loved reverse-engineering old games (Half-Life in particular) and restoring the original looks and properties of the code.

In this project I tried to use LLMs extensively to reverse the code for the first time to bring it back to its original potential look. Frankly, it worked and the code is now available to the public.