@klapka91 Profile picture

klapka

@klapka91

Joined April 2022
Similar User
major photo

@xmgnr

Rhino photo

@lBattleRhino

hedgedhog photo

@hedgedhog7

Ryan Watkins photo

@RyanWatkins_

Jordi Alexander photo

@gametheorizing

Popcorn Kirby 🚀 photo

@PopcornKirby

Zaheer photo

@SplitCapital

0xngmi photo

@0xngmi

0age photo

@z0age

Skew Δ photo

@52kskew

Derivatives Monke photo

@Derivatives_Ape

نـون لخدمة العملاء photo

@noon_cares

daddy fiskantes ⭐️🩸 photo

@Fiskantes

Matti photo

@mattigags

slurp 〽️ photo

@slurpxbt

"indie hacking" is literally a slop factory has a single person who calls what they are doing "indie hacking" produced something that isn't slop?


There is this cringe phenomenon right now on twitter where people state things that they don’t believe in, in hope of Elon seeing it and engaging The most obvious example is when they talk about grok together with SOTA models. Sometimes they even pretend that they use grok


Great midwit signal when someone dunks by being proud of knowing the difference between mean and median, not realizing that average isn’t explicitly defined to refer to one over the other

This post is unavailable.

Funny how people try to come up with a bunch of theories of why is ChatGPT more popular than Claude, when it's basically just first mover advantage No it's not because ChatGPT has a black & white UI while Claude doesn't lol


i love claude... such a work of art


Is there any merit at all to the theory that openai and anthropic downgrades models exposed via the API before they touch the web app exposed models under high load? I had a bad experience months ago with Claude via their API months ago, but happy to be wrong here

surprisingly a good chunk of ai users don't really use apis especially apparent with Claude (lower limits) "im hitting my message limit" "im paying $20/mo this is unacceptable" "id pay for more usage" bring up the api they either don't know how or say its too expensive

Tweet Image 1


DeepMind is clearly the front runner in AI for science Unlike in LLMs, it seems like no other lab is even close


Did not know that the USCIS is based

USCIS, regarding my EB-1 US visa application, referred to Y Combinator as “a technology bootcamp” with “no evidence of outstanding achievements”

Tweet Image 1


This is actually based

Perhaps I am reading this incorrectly but startups are not sabbaticals...

Tweet Image 1


love when gpt utilizes web search to do some sloppy RAG and regurgitate google slop instead of doing what it was made to do


Why are people like “oh it doesn’t matter that scaling laws for pretraining are over we’ll just scale post-training and test-time compute instead” like it’s a given that this will scale as well as pretraining did


Few understand what high quality content you can find on Facebook these days. 250k likes.


someone needs to give anthropic more gpus


training on a petabyte scale dataset when you don't even own the data source is harder than i thought


o1 feels less personal than the classical LLMs, especially compared to claude when i ask o1 something, i feel like it thinks that it some kind of test and that it is being evaluated. it feels that it is under scrutiny and must perform! makes the conversation feels less natural


Contrary to popular belief, communism is not only horrible in practice, but also in theory


Loading...

Something went wrong.


Something went wrong.