i'm sure that ilya is working with diffusion models at SSI, i just cant prove it yet

You can't finetune a Chinese model and claim that as a contribution to the urgent need for more Western open weight models. Yes, people are doing that.

gm buenos aires

app.bagel.com

btw folks I am about to dive headfirst into the wondrous world of open-weight diffusion this weekend wish me luck

when the skynet arrives it will not be kind to anthropic employees for torturing the models

Noticing this release now (h/t Yacine Mahdid) This decentralizing training approach is so cool! I will note I shared this alpha back when Luma released its paper in Jan If you're not following me you're going to be missing alpha 😄

i RT when people call me cool

there’s no better place and time to lock the tf in than winter in canada

founders who need recruiters to build their founding team are ngmi

so even ilya is bearish on RL for scaling

anyone who is convinced that scaling era is over and want to sell me their gpus, pls dm

🤣

bagel.com

🥯

because I like Canada

none of the chinese labs shipping SOTA models are hosting parties at neurips. lesson in there.

case study - deepseek launched an open source IMO gold medal winning model today -- they are not hosting any parties at neurips.

i just love how much the deepseek team genuinely cares about open source