petu

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 bundled HTML/JS being changed.

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 888 fiasco, but apparently it extended to 8 Gen 1 in OP10)

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 DownLoad) and files for it getting consistently "leaked". So at least exploratory work on custom ROMs should never leave you with a irrecoverable brick.

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 hot summers

Sounds like not a phone problem -- very high screen brightness and/or direct sun would make any phone hot.

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, then current day OxygenOS is just rebadged ColorOS. They just wont be pretending it's different now.

Only question/risk I see is Oppo trying to kill bootloader unlocking with an update.

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

packing multiple trits together

e.g. 5 trits (243 states) into a byte gives 1.6 bits per trit: https://compilade.net/blog/ternary-packing

Human could write that many comments to get enough base64 text for a design. Maybe to even get some of the highlighted characters in places they want (roughly equally spaced apart).

Indeed, I had a brain fart. Thank you.

But doesn't that make situation even worse? They likely would be 100B+ market cap/total asset value if spun off into independent company, but not able to generate even 1B in annual profit? Activision Blizzard managed 1.5B income/20% margin by themselves before being bought by Xbox -- and somehow whole of Xbox now earns less?

While it can be disregarded just as major update CS2 is formally 2023.

Dota is live service game, getting major patches each year. Patch notes for version released 3 months ago: https://liquipedia.net/dota2/Version_7.41 I don't think there's any reason for Dota 3, ever. Any change could be just a patch.

3% is lower than buying US Treasury bonds? Effectively zero risk zero effort return vs running a business.

https://home.treasury.gov/resource-center/data-chart-center/...

Nvidia reuses numbers for workstation cards, so there's multiple vastly different 'RTX 6000' cards:

  Quadro RTX 6000 (Turing / 2080 Ti / 24GB @ 672.0 GB/s)
  RTX A6000 (Ampere / 3090 / 48GB @ 768.0 GB/s)
  RTX 6000 Ada (4090 / 48GB @ 960.0 GB/s)
  RTX PRO 6000 Blackwell (5090 / 96GB @ 1.79 TB/s)

Maybe I didn't put it clear enough: __just the first__ GrapheneOS requirement leaves us with three brands. There's 20+ more items on the list after that, Nothing and Fairphone likely fail on them.

Pixels are consistently "third party Android builds friendly", plus GrapheneOS has a list of required security features (beyond their control): https://grapheneos.org/faq#future-devices

e.g. first one in the list:

Support for using alternate operating systems including full hardware security functionality

GrapheneOS wants users to lock the bootloader (≈enable Secure Boot) after install by providing user signing keys (avb_custom_key) -- that already seems to leave only Pixel, Nothing and Fairphone.

https://github.com/chenxiaolong/avbroot/issues/299

How's that relevant?

There's no Steam subscription of any kind to access your library. Whether you're spending $0 on average year, or $300 -- you get to access your Steam library the same.

There's talks about HBF. F for Flash -- HBM packaging and bus width, but using NAND memory.

e.g. https://www.sandisk.com/company/newsroom/press-releases/2026...

That's attribution to Qwen team.

There (is/was) no attribution to Nex team (they've released a model based on Qwen 3.5 397B as well).

As per OP link Nex claims that what Rio team released (so far) is just linear interpolation of weights between Nex and OG Qwen model. With no attribution to Nex and zero signs of Rio doing any training of their own.

You're kinda talking about Claude being used for planning/architect role, while local LLM is just executing it (performing edits) -- at least in such form it's a thing, yes.

I can imagine it being ten or hundred times more

I don't think I average even 2 captchas a day being terminally online, so 10 across every soul in the world sounds way too much for me. (we're ignoring bots it's meant to deter?)

it's hard to see that such a modest energy cost would dissuade any attacks.

It's not against targeted attacks, but scrapping.

And not about energy cost, but available compute power -- it requires scrapper to use browser with JS (or time commitment to reimplement PoW outside of JS), limits their request rate by CPU core count.

but unless you do PoW (which is also ecologically a nightmare)

Can you expand? I don't see a problem with some napkin math. 5W load for 2 seconds is 0.002Wh (we have to let smartphones pass and not by doing PoW for 10s of seconds). 8 billion checks a day for a year = 8GWh.

Owned by Lenovo, yes

Note that Xiaomi today is very hostile to bootloader unlocking -- to get unlock code you need to win in daily first come, first served "lottery" (they limit total number of unlocked devices per day). After a second or two passes you're already too late for the day.

Half of the world lives on $300/mo. For majority of the world there's meaningful impact in saving $20 on a book.

It's digital copies, no real damage is done.

AA asks you to not scrape them because of server load and provides torrents to download everything in more efficient manner.

How do you know that Anna's archive started operating in 2022?

edit: you've sent me Wikipedia link and then removed your reply. So I'll put my reply here:

https://en.wikipedia.org/wiki/Anna%27s_Archive

Very first sentence in article:

Anna's Archive is an open source search engine for shadow libraries that was launched by the pseudonymous Anna shortly after law enforcement efforts to shut down Z-Library in 2022.

Doesn't it clearly say that there's 'prior art'? So much so, that there's dedicated 'shadow library' article linked?

With that basic context (you should've been aware of?) your speculation makes zero sense:

But perhaps it was set up by AI training thieves. The founding date of July 2022 would speak for that theory.

formal Russian is 'em dash heavy', here's South Korea Wikipedia arcticle in Russian, Dec 2021 version: https://ru.wikipedia.org/w/index.php?title=%D0%A0%D0%B5%D1%8...

Ignoring references, just in article text: 5800 words (spaces); 78 em dashes (1.3%); 0 en dashes; 90 hyphens (1.5%).

English version of same timeframe: 16000 words; 0 em dashes; 32 en dashes (0.2%); 262 hyphens (1.6%).

The emojis used in the bullet points

They're used in one list, where sub-projects are listed. Emojis used in that list are consistent with ones used for same sub-projects on Wiki https://docs.flipper.net/one Someone thought it would be better than plain text, that's it.