Unfortunately this is not true and viruses use your body to mine shitcoins and record some random crap on the chain (your DNA) I bet if you carefully look at the “junk DNA” regions you will find jpegs with ugly monkey pfps

The brother of my Georgian grandfather was Giorgi, but everyone called him Yuri (what?) I suspect Muslims have a similar mechanism to distinguish various Mohameds

There are some outlier cultures where it is customary to give everyone the same name, and it gets weird In Georgia a huge proportion of the male pupulation share the name Giorgi, so they come up with dozens of diminutive versions of Giorgi to distinguish (Gogi, Gegi, Giga, etc)

Children’s names lie on a lame-cringe spectrum If a name is very popular it is lame, if a name is rare and exotic it is cringe. Parents have to pick a sweet spot somewhere in the middle, which results in a slow but steady distributional drift

Can anyone explain why there is so much hype about Omarchy? This is just another Arch derivative, no?

The entire architectural complex is a huge Red Alert style neo-Stalinist Disneyland You need to understand the context

Western larpers don’t understand that the first building is a postmodern architectural shitpost and has nothing to do with Orthodoxy It is brilliant but for all the wrong reasons The other two are just ugly

If you are participating in a public debate where your assumed goal is to convince a diverse crowd of people, why would you ever base your argument on your personal religious beliefs? This implies either 1) Stupidity 2) Convincing anyone was never a goal

Very weird to see companies doing traditional B2B SaaS adopting the “member of technical staff” job titles Cargo cults are everywhere

The age of slopware is here

Gotcha!

When you open a website and immediately notice the suspiciously familiar layout, elements, icons, shapes of buttons and subtle text misalignment issues

It was, in fact, degenerate gambling

Can you find a cat?

I want to live in a world where companies like Cluely don’t survive

It is purely speculative of course, but it may be a plausible explanation of why @thinkymachines spend so much effort on building LoRA infra for example Would love to see some research on LoRA compression efficiency

So it seems like for longer context lengths it may be reasonable to spend a large amount of compute to squeeze the information out of the KV cache and move is somewhere else

From @jxmnop research we know that LLMs compress information to around 3.6 bits/parameter in their FFN layers In comparison, KV cache only stores ~1e-4 bits (!!!!!) per parameter Looks extraordinarily wasteful!