Hacker News: 5555watchhttps://news.ycombinator.com/user?id=5555watchHacker News RSShttps://hnrss.org/hnrss v2.1.1Sat, 09 May 2026 16:47:11 +0000<![CDATA[New comment by 5555watch in "A recent experience with ChatGPT 5.5 Pro"]]>I would guess it's because ChatGPT Pro allows for 80min "think". I've never had even remotely similar think times with Gemini Deep Think. It's generally around 10-15min for math problems, and get increasingly shorter for continued interactions.

]]>Sat, 09 May 2026 10:20:25 +0000https://news.ycombinator.com/item?id=480737245555watchhttps://news.ycombinator.com/item?id=48073724https://news.ycombinator.com/item?id=48073724<![CDATA[New comment by 5555watch in "Google releases Gemma 4 open models"]]>Why no (high) variants in the comparison models?

]]>Fri, 03 Apr 2026 14:33:48 +0000https://news.ycombinator.com/item?id=476271065555watchhttps://news.ycombinator.com/item?id=47627106https://news.ycombinator.com/item?id=47627106<![CDATA[New comment by 5555watch in "TinyLoRA – Learning to Reason in 13 Parameters"]]>It's not "13 parameters to reason", they just rotated the full 8B parameter space in 13 dimensions and found a rotation that was still able to reason.

Depending on the latent structure, it's possible a nice rotation that would be perfect for some one specific problem, but you still got to search for it, and it's not a guarantee to exist.

But it's a nice step towards LLM parameter-space interpretability.

]]>Thu, 02 Apr 2026 01:59:31 +0000https://news.ycombinator.com/item?id=476091655555watchhttps://news.ycombinator.com/item?id=47609165https://news.ycombinator.com/item?id=47609165