Playground AI founder Suhail teases a new AI venture equipped with two 8x Nvidia B200 systems
Garry Tan publicly supported the unrevealed AI project.
Many users express excitement about the founder restarting the AI project with dual 8x B200 GPU systems because it empowers solo builders by compressing months of learning into hours.
Most Activity
@Suhail LFG
1/ it all started w 2 8xB200s
excited to be back in the game again
3/ Time to let my autonomous ai scientist rip on some new optimizations
2/ spent a lot of time reviewing the absolute fundamentals again; missed a lot in the world while working on image models
there’s so much out there / so many talented people to learn from
@Suhail I remember you posted this for MightyApp.. something like… It all started with “60ms”
2/ spent a lot of time reviewing the absolute fundamentals again; missed a lot in the world while working on image models
there’s so much out there / so many talented people to learn from
1/ it all started w 2 8xB200s
excited to be back in the game again
@Suhail Pumped
1/ it all started w 2 8xB200s
excited to be back in the game again
@Diveej_
@Suhail @garrytan Time to build (a generational company).
1/ it all started w 2 8xB200s
excited to be back in the game again
@moonlght12891 not an image model 🫣
@Suhail Hey Sushil. why did you shut down that VM company??
@garrytan appreciate you brother
@Suhail just fyi, they're nerfing perf on AI research tasks
DO NOT USE FABLE 5 FOR AI R&D/CODING!!
@Suhail 🤜🏼🤛🏼
@Suhail Any hints on what you are training
@_mohansolo 🙏🏾♥️🫡
@Suhail It really is a superpower. Being able to compress months of learning into hours changes the whole game for solo builders. Looking forward to the updates!
@iScienceLuvr I shall test :)
@Suhail just fyi, they're nerfing perf on AI research tasks
@kevinwooof 🫡
@olivercameron @garrytan 🫡
@Suhail should I buy a B300 pod?
i'd kind of prefer 4x gh200 dualnode racks, can add more later ... really not sure ...
@Suhail the inference margin story is brutal right now. B200 only flips the game if you need sub-100ms latency or density was somehow the blocker all along. which is it?