xiaoyu2006

The blog write-up style is so casual haha.

I assume different models will have different distribution, so it has to be kept updated?

Not quite worth its high price

They don't sell well in Asia. It's mainly xiaomi, oppo, vivo and huawei.

Ownership is a very limiting coding constraint, and bypassing methods, when ultimately needed (e.g. cell), cannot be engineered to be ergonomical.

Not playing the politics game is also a political stances. There is no way to avoid it.

Corporation machine are not individuals (geeks). They chase solely profit while we have (at least some) agency on supporting the moral sides.

(Just fyi the correct answer is that organizer must report the organizing beforehand or it would be illegal.) Model labs have to censor their models in order to publish them, which is not equlv to model lab management members actively showing their political stances.

I always wonder why use a DSL while you can just borrow some existing PL?

...for fun!

Bad idea. It either takes a useless subset of all languages or do weird abstraction that wouldn't fit in a language otherwise.

...recursion?!

It should not. Abstraction in software engineering brings intelligence. (compression correlates to intelligence)

this is such a good laugh. peak comedy

Some models are heavily subsidized. Total params & active params are better measurement of inference cost.

This open source model is quite near SOTA with only 700B/40B MoE. Truly efficient.

V100 32G SXM2 adapted to PCIE. Running llamacpp with Q4KM Qwen 3.6 27B or Gemma 4 31B. I use them when I feel of privacy is important or I just want to mess around.

Hmm I wonder why one gets attention and the other did not. HN need the "duplicate" feature SO had.

Yes

Obligatory: https://github.com/chardet/chardet/issues/327

How is Magic Mouse even close to ergonomics lol

Reminds me that when I was in high school, my father used to regulate my gaming sessions with Apple ScreenTime. I changed the BundleIdentifier in my game's info.plist to `com.apple.systempreferences` and sign the .app with my own developer cert to bypass this. Was a lot of fun before got caught physically.

I hope they redesign their magic mouse. It's not a real product.

I always think 401k is not fair at all. It kinda forces one to invest and pump the stock prices.

These companies are too big to fail. I'm afraid the tax payers will be the ultimate consequences bearer.

I was expecting a post on how hardware accelerate linear algebra lol

Even with that, 1G of RAM usage is still not justified.

Yeah the weird notation confused me too. Their own Limitations also says their experiments are too small. I am quite curious how it will play out big now, but unironically I cannot afford the hardware lol.

Will be great and amusing if it actually turns out that we have been doing transformer overly-complex. The code repo is missing tho...

Exactly. Repeating or rephrasing a definition is trivial, teaching someone is not.