hham
Re:Samsung's Processing-in-Memory (PIM)
on Saturday August 29, 2026 @07:07AM (#49488901)
memory plus memory bandwidth > GPU, this is the gist of it.
Re:American AI is locked down and proprietary. It's losing
on Monday July 20, 2026 @03:10PM (#48983481)
Totally agree, we wrote an article about ho this is a senseless "red queen race" few days ago, hope is ok to post: https://news.ycombinator.com/item?id=48892559
Re:Inkling: Our Open-Weights Model
on Thursday July 16, 2026 @06:50AM (#48932793)
Re:Inkling: Our Open-Weights Model
on Thursday July 16, 2026 @06:17AM (#48932563)
Smart that she says "not the strongest overall model, open or closed". This is a rare for an AI lab to say out loud. They basically decided to compete on customizability, and not on topping the temporary leaderboard. Also corroborates what we recently wrote: any Lab's capability lead cant hold for long anyway, it's a "red queen race" that never settles: https://news.ycombinator.com/item?id=48892559
Re:Bonsai 27B: A 27B-Class Model that runs on a phone
on Wednesday July 15, 2026 @12:38AM (#48916350)
This is accelerant #3 and #4 from our article converging in one release: a 27B-class model, built on Qwen (already one of our examples of local models "good enough to matter"), now running on an iPhone. The hardware layer and the local-model layer aren't just going to converge in the future, they're doing it right now! https://news.ycombinator.com/item?id=48892559