Jj McMc

[Jj McMc @AccountMus629]

they pay me to post here

[Jj McMc @AccountMus629]

@tszzl LoC is a useless metric when it is a metric you are specifically aiming for or when talking about multiple languages, if you work on similar stuff and no one purposefully manipulates LoC count, it is actually pretty comparable

[Jj McMc @AccountMus629]

@jacobsali123 @stupidtechtakes Yeah, again, an attacker can just check if you are already signed up on the signup form, on like 90% of pages ive ever checked

[Jj McMc @AccountMus629]

@scaling01 I just dont see why that would be the case, like, wouldnt openai just train a much larger model if training took that little compute?

[Jj McMc @AccountMus629]

@vaxryy use hyprland

[Jj McMc @AccountMus629]

@merobbins @petergostev In the codex app for macos, settings -> profile

[Jj McMc @AccountMus629]

@mhdcode Hmm usually i would agree, but here, idk. Anywhere where 3.1 pro or flash are at the top - totally, absolutely true. But 3.5 flash scoring between kimi and qwen is honestly pretty realistic, it is a bad model for google, but it isnt completely trash

[Jj McMc @AccountMus629]

@ml_angelopoulos LMarena is shit

[Jj McMc @AccountMus629]

@cnakazawa But at least you now get to say a weird acronym every time, plus your coworkers will all just call it "Chat", which will drive you up the wall

At least that is what chat told me

[Jj McMc @AccountMus629]

@quadcodeai @henokcrypto Oh wow, brilliant insight, next youre going to tell me what matters about cars is how they drive and not their name

[Jj McMc @AccountMus629]

@haider1 Thank you for saying something everyone already knows, again and again

[Jj McMc @AccountMus629]

@BLiangAA @thsottiaux Thank you for this useless trash graph, you are really showing that you dont have a fucking clue

[Jj McMc @AccountMus629]

@Everlier @iruletheworldmo Thanks for your fucking useless contribution, at least you filled the internet with a few more bytes of needless bullshit

[Jj McMc @AccountMus629]

@iruletheworldmo Dude just shut the fuck up, please

[Jj McMc @AccountMus629]

@birdabo that guy keeps saying a lot of things and like 10% are true, search for 5.6 on his profile, he already claimed like 3 times that it would release the next day and it didnt

[Jj McMc @AccountMus629]

@JasonBotterill I dont think it is, its not like cerebras chips got large since last time

[Jj McMc @AccountMus629]

@crypt0lake I mean you cant really FORCE a model to reason a lot, if the user says "Hi, can you help me with a question" reasoning for 16000 tokens would be a waste of time

[Jj McMc @AccountMus629]

@joko76ers @scaling01 Yeah i mean this benchmark wasnt designed to be realistic, it was designed to be hard AF for llms

[Jj McMc @AccountMus629]

@scaling01 Extrapolating from only these 2 data points and with a method i wont publish, i have determined that we will have AGI by thursday, june 4, 2026

[Jj McMc @AccountMus629]

@wordgrammer id rather be unemployed than write frontend