Hacker News: osti

New comment by osti in "Fuck You, Bambu Lab. Go Ahead, Sue Us"

osti — Wed, 13 May 2026 04:11:16 +0000

Reality has both negatives and positives. Gamers Nexus clearly lie on the other side of the spectrum where they overwhelmingly choose the negative stuffs, hence rage baits.

New comment by osti in "Fuck You, Bambu Lab. Go Ahead, Sue Us"

osti — Wed, 13 May 2026 02:29:24 +0000

I understand their points. But to me they do too much rage baits these days that I can't bring myself to watch their stuff. Why would I let my self get angry lol.

New comment by osti in "Bun's experimental Rust rewrite hits 99.8% test compatibility on Linux x64 glibc"

osti — Sun, 10 May 2026 03:54:35 +0000

Even last year at this time people wouldn't believe it.

New comment by osti in "ProgramBench: Can language models rebuild programs from scratch?"

osti — Thu, 07 May 2026 06:24:11 +0000

I heard they are already proficient at assembly languages.

New comment by osti in "Show HN: Auto-Architecture: Karpathy's Loop, pointed at a CPU"

osti — Wed, 29 Apr 2026 06:20:16 +0000

A lot of it was beyond me, but this was all the branch names for all the stuff it tried, most of it unsuccessful of course. About 10x perf improvement came from architectural changes, and then 2x from micro optimizations.

https://pastebin.com/eac0SAYg

New comment by osti in "Show HN: Auto-Architecture: Karpathy's Loop, pointed at a CPU"

osti — Wed, 29 Apr 2026 04:30:44 +0000

> propose, implement, measure, keep the wins

Pretty much what I did to let Codex with gpt5.4xhigh improve my fairly complex CUDA kernel which resulted in 20x throughput improvement.

New comment by osti in "SWE-bench Verified no longer measures frontier coding capabilities"

osti — Mon, 27 Apr 2026 01:55:04 +0000

This was published two months ago. Even though it was at a time that open source models are publishing comparable swe bench scores.

New comment by osti in "GPT-5.5"

osti — Thu, 23 Apr 2026 22:09:35 +0000

Can't they write a script to solve rubik cubes?

New comment by osti in "Kimi K2.6: Advancing Open-Source Coding"

osti — Mon, 20 Apr 2026 18:52:10 +0000

True, but I think for local models, we are mostly considering personal usage.

New comment by osti in "Kimi K2.6: Advancing open-source coding"

osti — Mon, 20 Apr 2026 18:50:54 +0000

Yeah... I would definitely call 2t/s unusable. For simple chats, I'd want at least 15 t/s. For agentic coding (which this model is advertised for), I'd want good prefill performance as well.

New comment by osti in "Kimi K2.6: Advancing Open-Source Coding"

osti — Mon, 20 Apr 2026 17:06:46 +0000

Oh i’m fully aware of that lol

New comment by osti in "Kimi K2.6: Advancing open-source coding"

osti — Mon, 20 Apr 2026 16:33:42 +0000

I think this one is only about 600GB VRAM usage, so it could fit on two mac studios with 512GB vram each. That would have costed (albeit no longer available) something like less than 20k.

New comment by osti in "Kimi K2.6: Advancing open-source coding"

osti — Mon, 20 Apr 2026 16:31:25 +0000

Maybe open source == communism

New comment by osti in "Taste in the age of AI and LLMs"

osti — Sat, 11 Apr 2026 04:46:52 +0000

I can't vouch for whether or not it can beat human experts though because I'm no CUDA expert myself. The original CUDA code were human written and I first let codex adapt it to my specific use case. Then I basically let codex generate ideas and try the ideas out itself (I think it's a bit like Karpathy's autoresearch, except I was still doing manual prompting). And that was enough to get me 20x improvement.

I suspect when people said AI wrote non performant CUDA kernels it was beginning-mid last year and it's definitely vastly improved since back then. And the agent's ability to iteratively improve really impressed me.

New comment by osti in "ChatGPT Pro now starts at $100/month"

osti — Sat, 11 Apr 2026 00:54:18 +0000

I was using codex cli with 5.4xhigh. So it was able to iteratively improve from simple prompts on my part (can you give some architectural ideas to improve the performance? And once it does, I just say can you implement and benchmark it).

I think it was a bit like Karpathy's autoresearch, except I was doing manual promoting... Though I feel I could definitely be removed from that equation.

New comment by osti in "Taste in the age of AI and LLMs"

osti — Thu, 09 Apr 2026 23:38:10 +0000

For me it was able to try out different architectures for perf improvement, then once it's settled on some good architectures, it can do lower level optimizations on them by profiling the code etc.

New comment by osti in "ChatGPT Pro now starts at $100/month"

osti — Thu, 09 Apr 2026 19:59:28 +0000

Yup I've mentioned this in another thread, I got gpt 5.4xhigh to improve the throughout of a very complex non typical CUDA kernel by 20x. This was through a combination of architecture changes and then do low level optimizations, it did the profiling all by itself. I was extremely impressed.

New comment by osti in "Taste in the age of AI and LLMs"

osti — Tue, 07 Apr 2026 17:44:10 +0000

I had a very complex cuda kernel and codex cli managed to improve the throughout 20x.

New comment by osti in "Google releases Gemma 4 open models"

osti — Thu, 02 Apr 2026 19:16:26 +0000

But is arc-agi really that useful though? Nowadays it seems to me that it's just another benchmark that needs to be specifically trained for. Maybe the Chinese models just didn't focus on it as much.

New comment by osti in "Slop is not necessarily the future"

osti — Wed, 01 Apr 2026 01:16:50 +0000

That is ture, but the revenue of the artisanal stuff is probably only a very low percentage of the overall market, which would imply a lot of software engineers would have to exit the field. Which is what we here don't want to see.