Engin Kaya

[Pablo Fredrikson @PeladoNerd]

https://t.co/zfFpPnJbdr

[Борис Трушин @TrushinBV]

Немного физики )

Вы замечали, что на большинстве иллюстраций айсберга его рисуют неправильно? Длинная глыба льда не будет плавать вертикально

[Nav Toor @heynavtoor]

THIS GUY BUILT AN ENTIRE WIKIPEDIA THAT IS 100% AI HALLUCINATIONS AND IT'S OPEN SOURCE ON GITHUB

it's called Halupedia.

nothing on the site existed before you clicked. every article was generated the second you arrived.

the site has one rule: the universe only exists when you visit it.

it looks exactly like wikipedia. same fonts. same layout. same scholarly citations. same "stumble" button for random articles.

the only difference is none of it is real.

here are some actual articles currently in the encyclopedia:

> the great pigeon census of 1887
> the ministry of slightly wrong maps
> chaldic arithmetic — a branch of mathematics where subtraction is forbidden
> armund the river mapper — a cartographer who mapped 14,000 leagues of river without leaving his chair
> the society for the prevention of unnecessary tuesdays

every article page also tells you how many people are reading it right now. it says: "you alone are consulting this folio at present."

the creator's own tagline for the site is the most unhinged sentence i've read this year:

"an encyclopedia of a universe that does not exist until you visit it"

the entire backend is a single open source repo called vibeserver. one guy. one description on github: "a little webserver making things up just in time."

we built the largest knowledge base in human history and the very first thing a guy did with it was make a hallucinated mirror universe and put it on the open web.

the internet is healing.

[Andrej Karpathy @karpathy]

Software horror: litellm PyPI supply chain attack.

Simple `pip install litellm` was enough to exfiltrate SSH keys, AWS/GCP/Azure creds, Kubernetes configs, git credentials, env vars (all your API keys), shell history, crypto wallets, SSL private keys, CI/CD secrets, database passwords.

LiteLLM itself has 97 million downloads per month which is already terrible, but much worse, the contagion spreads to any project that depends on litellm. For example, if you did `pip install dspy` (which depended on litellm>=1.64.0), you'd also be pwnd. Same for any other large project that depended on litellm.

Afaict the poisoned version was up for only less than ~1 hour. The attack had a bug which led to its discovery - Callum McMahon was using an MCP plugin inside Cursor that pulled in litellm as a transitive dependency. When litellm 1.82.8 installed, their machine ran out of RAM and crashed. So if the attacker didn't vibe code this attack it could have been undetected for many days or weeks.

Supply chain attacks like this are basically the scariest thing imaginable in modern software. Every time you install any depedency you could be pulling in a poisoned package anywhere deep inside its entire depedency tree. This is especially risky with large projects that might have lots and lots of dependencies. The credentials that do get stolen in each attack can then be used to take over more accounts and compromise more packages.

Classical software engineering would have you believe that dependencies are good (we're building pyramids from bricks), but imo this has to be re-evaluated, and it's why I've been so growingly averse to them, preferring to use LLMs to "yoink" functionality when it's simple enough and possible.

[critter @BecomingCritter]

brilliant

[Engin Kaya @kaengn]

https://t.co/m2PC7bU4wM

[Rohan Paul @rohanpaul_ai]

Good GPU performance summaries - in 6 mints.

[Engin Kaya @kaengn]

https://t.co/fOr5Z3hLf3

[Engin Kaya @kaengn]

Attention? Attention! https://t.co/eUpXSvYCl5 #architecture #attention #transformer #rnn

[Engin Kaya @kaengn]

https://t.co/yyQ6uYKIj0

[Engin Kaya @kaengn]

@sam_synnada This talk was good https://t.co/FgQlDobXkM

[Engin Kaya @kaengn]

Podcast Notes /// Richard Sutton – Father of RL thinks LLMs are a dead end | Dwarkesh Podcast https://t.co/V7dr5oTg84

[Engin Kaya @kaengn]

https://t.co/ltOs5fMAhx

[Oumar Kaba @sekoumarkaba]

Peer review is dead

[Engin Kaya @kaengn]

https://t.co/flddOMhigW

[Engin Kaya @kaengn]

https://t.co/ykJqC1uyi1

[Sara Hooker @sarahookr]

Truly excellent video by @MLStreetTalk about how a handful of providers have systematically overfit to @lmarena_ai.

26 mins of video showcase how easy it has been to distort the rankings.

As scientists, we must do better. As a community, I hope we can demand better.

[Engin Kaya @kaengn]

How language model post-training is done today https://t.co/juPkMJIrRe via @YouTube

[Engin Kaya @kaengn]

@akshay_pachaar Right, great point! Thank you