Ram

scared of @sama

[Ram @ram_devv]

@evilmathkid @arcprize @fchollet @mikeknoop ye i have been working on adaptive attention on dsa and curious if looped ones use just softmax or smth custom

[Ram @ram_devv]

@evilmathkid @arcprize @fchollet @mikeknoop just a curious question have you worked on attention?

[Ram @ram_devv]

@evilmathkid @arcprize @fchollet @mikeknoop this is so cool!

[Ram @ram_devv]

@theo can you add STT for the mobile app?

[Ram @ram_devv]

@zachmoskow @JensenHuang lol

[Ram @ram_devv]

@Suffix_AI @rezoundous buddy we are about the same age, im mad cause of ur lack of intelligence to understand and fun fact i have a job unlike your unemployed ass.

[Ram @ram_devv]

@Suffix_AI @rezoundous 30% of the evals is their own research, it is "a benchmark", do your research before talking smth stupid.

[Ram @ram_devv]

fogot to mention hyprland.....

[Ram @ram_devv]

omarchy tiling + the glass plugin + foot feels good

[Ram @ram_devv]

anybody got ideas for new attention method for llms ?

[Ram @ram_devv]

@Suffix_AI @rezoundous also if you trust a benchmark where google is at the top, you are one stupid mf.

[Ram @ram_devv]

@theo @AdamHoltererer excited to see what you do with the data!!

[Ram @ram_devv]

@Suffix_AI @rezoundous we are talking about opus on low reasoning browski

[Ram @ram_devv]

@aarondelasy dude i agree
it feels longer in t3 idk why?

[Ram @ram_devv]

lol

[Ram @ram_devv]

@orhundev i think i've heard this one before.

[Ram @ram_devv]

@ArifHus18238070 @nanaaji4440444 lol

[Ram @ram_devv]

@rezoundous literally every benchmark you see haiku on high beats opus on low

[Ram @ram_devv]

@rezoundous dude haiku high > opus low

[Ram @ram_devv]

@Suffix_AI @rezoundous i hate kids who dont understand the point of reasoning levels.

[Ram @ram_devv]

@OpenAI they just removed some neural layers decensored 6 sol and called it cyber.