serrano_hux

@NVIDIAGeForce #rtxon

Stand and Fight for a chance to win. We're giving away a #DOOM: The Dark Ages custom wrapped NVIDIA GeForce RTX 5090 to celebrate its official release with native ray tracing, NVIDIA DLSS 4 + Reflex. To enter: 🟢 Like this post 🟢 Comment #RTXON

@PuppyyPawss I already parked my car

dream about me

@ZVR09 Massive deficits

Replace IRS with flat tax. Everyone pays the same rate above 50k. Your first 50k is not taxed, over 50k is taxed at 15%. Zero deductions, zero loopholes. Tax form is 4 lines:

@tekbog Mofo won’t leetcode talking about existential battles

i dont want to heal i want to be put in a battle to finish me or ascend beyond what i have ever been capable of either way i will be free

naming your child is the single greatest leverage point for determining their future, certainly much higher than choosing what $60k/y brooklyn heights preschool to send them to choosing based on a popular name list is a guarantee to live life at median, escape

@tekbog lol same bro

i keep gaslighting the google recruiter every month by telling her ill eventually take the interview

i want to get a new job but i dont wanna grind leetcode it's so fucking over for me

@Andr3jH Beware of a man who talks about love but has never fucked.

the only person who would have actually benefitted from a good 6D mesoamerican gigafrying

@buccocapital It’s far more family friendly and as millennials are now in child raising years it’s inevitable

The boomers are mounting one last push for in-office work Looking pretty bleak for WFH over the next five years But the trend/technology is inevitable. A few years from now it’s worth throwing some big chips down on WFH as the future

@AllanLichtman Keys open doors

Nate Silver has finally seen the light! Weeks after I predicted a Harris victory he has come down from a 2/3 probability of a Trump victory to a 58% probability of a Harris victory.

@deliprao SageMaker studio has this functionality.

is there an llm finetuning service that will accept my data, train an open model (say llama3.2), and allow me to download the trained model?

4 months ago I worked on A* for reasoning. Hypothesis on how O1 works: > actor model generates outputs at depth 1 (Monte Carlo style) > critic model scores outputs with a heuristic > model generates depth 2 > run a* search algorithm It is an actor critic style LLM

This may be the most important figure in LLM research since the OG Chinchilla scaling law in 2022. The key insight is 2 curves working in tandem. Not one. People have been predicting a stagnation in LLM capability by extrapolating the training scaling law, yet they didn't foresee that inference scaling is what truly beats the diminishing return. I posted in February that no self-improving LLM algorithm was able to gain much beyond 3 rounds. No one was able to reproduce AlphaGo's success in the realm of LLM, where more compute would carry the capability envelope beyond human level. Well, we have turned the page.

Today, I’m excited to share with you all the fruit of our effort at @OpenAI to create AI models capable of truly general reasoning: OpenAI's new o1 model series! (aka 🍓) Let me explain 🧵 1/

Akira

Haskell is the Scientology of programming languages

@micahjay1 Almost by its nature PE should outperform VC. VC is speculative while PE is not (or shouldn’t be).

an LP said to me today - "the median Private Equity firm way outperforms the median VC firm"

@Yampeleg So true

$TSLA massive outperformance last few days. Benefiting from the probable rate cut this month. I’m thinking Tesla ends up at $300 by end of year.

@petergyang @ctjlewis I don’t get why YouTubers are more important than ai researchers but idk whatever gets views. Scarlett Johnansson is much sexier than Hinton.

It's hilarious that Scarlett Johansson is pictured here, but there's no: Sam Altman Dario Amodei Ilya Sutskever Andrej Karpathy or even...Mark Zuckerberg