Viewing profile — petu

If algorithm used is static and GPU-friendly, then what stops bypass at scale?

This page has it, scroll to "Intelligence" header (not the highlights one, but second on the page / with black square) and click "Agentic Index"

Hm, I asked 5.6 Sol to calculate/research and it came up with way less optimistic lifespan than Gemini (that I've asked last year). It claims that each individual page read induces…

There's read disturb on SSDs, enough reads will eventually force controller to rewrite the cell and it's neighbours. Practically if you're not streaming weights 24/7 from a full SS…

Does Claude still say it's Deepseek, when asked? https://news.ycombinator.com/item?id=49082022#49087112 How is that important? Maybe it does, so what?

Weights are yet to be released (maybe in 24H, Deepseek has track record of releasing same day). https://github.com/antirez/ds4 is often mentioned for DS4F on Mac, but 64GB is likel…

It probably would be called 'deepseek-v4-flash-0731' in API edit: nope, at least deepseek kept "deepseek-v4-flash" and just updated model underneath. I guess preview is no longer w…

Maybe .sexy TLD owner decided that "onlyfans" is a "premium" domain, with special pricing. And registrar isn't aware of premium domain pricing (at least for that TLD).

Unused RAM is wasted RAM. So not really Apple magic, about every OS uses "free" memory as disk cache. Try to leave only a gigabyte or two free, speed likely would drop dramatically…

You don't get to take distilled model home, it all stays with Google. It's "optimize your costs in our garden" product.

LLMs can't reliably answer what they are (w/o getting that info from system prompt/tool call/etc), so yea. https://xcancel.com/teortaxesTex/status/2026130112685416881 I think I've …

They're not saying anything about Anthropic serving costs in that quote, just calculating what MTok price is for running agents. Next sentence after your quote: > Agentic workloads…

I think internal use is allowed at any scale in the license? > 4. The requirements set forth in Sections 2 and 3 do not apply to: (a) internal use of the Software, defined as any u…

You need whole weights in VRAM for optimal performance. Don't be confused by "experts" in the name -- you don't get to load static subset of experts and blast next 100 tokens with …

First sentence is about token efficiency.

I think in case of Firefox it's done purely with JS ( https://mozilla.github.io/pdf.js/ ), no native code. So that feels fine (assuming it runs in same sandbox like any page), just…

To be fair Deepseek V4 now is just a preview, which likely implies training isn't completed.

Deepseek V4 paper is just ~three months old

I've misread that you're having consistent problems across different models, sorry. If it's just 10 Pro, then google says Qualcomm was having bad years (I've heard about Snapdragon…

It's also 2.8x parameter count (1T -> 2.8T), likely higher activation per token (50B?).

Nowhere close to what you asked, but Nothing already seems to be one of the most "open" devices today. That's for simple fact of having offline/unsigned EDL (Qualcomm's Emergency D…

> it seems that it still runs hot according to some reviews I've seen on the yt. Never noticed it being even warm in normal use, consistently cold. > this is especially true during…

As other commenter said OnePlus is subsidiary of Oppo. Over past few years they were loosing autonomy / reusing more and more work by Oppo. If you're worried about the firmware, th…

Most people use official clients and AFAIK they don't really accept any external contributions? All clients were fully developed by employees.