Hacker News: 0xkvybhttps://news.ycombinator.com/user?id=0xkvybHacker News RSShttps://hnrss.org/hnrss v2.1.1Sun, 03 May 2026 16:22:20 +0000<![CDATA[New comment by 0xkvyb in "Kimi K2.6 just beat Claude, GPT-5.5, and Gemini in a coding challenge"]]>Totally agree with you. There is only so much time before SF tech runs out of subsidy bucks, and Chinese models take the consumer spotlight

]]>Sun, 03 May 2026 06:25:41 +0000https://news.ycombinator.com/item?id=479939690xkvybhttps://news.ycombinator.com/item?id=47993969https://news.ycombinator.com/item?id=47993969<![CDATA[New comment by 0xkvyb in "DeepSeek V4–almost on the frontier, a fraction of the price"]]>Yes, GLM 5.1 is surprisingly good! Particularly for long-horizon Agentic tasks, with 100+ available tools. It really shocked me in a good way when it was able to complete a long run with 50+ steps and not fall into a loop along the way.

]]>Sat, 02 May 2026 17:39:44 +0000https://news.ycombinator.com/item?id=479885790xkvybhttps://news.ycombinator.com/item?id=47988579https://news.ycombinator.com/item?id=47988579<![CDATA[New comment by 0xkvyb in "DeepSeek V4–almost on the frontier, a fraction of the price"]]>It might be at the frontier, but DeepSeek is really struggling with compute. The amount of 429 Rate Limit responses I've been getting just testing this thing made me pause all my attempts at cross-comparing it to others.

I'm gonna stick to GLM5.1 for now.

]]>Sat, 02 May 2026 17:37:20 +0000https://news.ycombinator.com/item?id=479885500xkvybhttps://news.ycombinator.com/item?id=47988550https://news.ycombinator.com/item?id=47988550<![CDATA[New comment by 0xkvyb in "Refusal in Language Models Is Mediated by a Single Direction"]]>Still crazy how easy it is to "jailbreak" even SOTA LLMs with a simple assistantResponse replacement in chat thread.

]]>Sat, 02 May 2026 17:32:29 +0000https://news.ycombinator.com/item?id=479885050xkvybhttps://news.ycombinator.com/item?id=47988505https://news.ycombinator.com/item?id=47988505