23, kernel dev, mle @synclabs, fuck it we ball
Joined August 2019
- Tweets2.4K
- Following1.5K
- Followers706
- Likes29.4K
Pinned Tweet
hand wrote a blackwell gemm in cutedsl and went from 164 to 1401 TFLOP/s on a B200, about 97% of cuBLAS
back home from sf, made quite a lot of good memories. now we start grinding so that we can reach our goal much faster ;)
expect a lot of tweets regarding model optimization from an inference and training perspective.
LFG!!!!
Vishal retweeted
in September, we took over the biggest media conference in Europe.
sync. labs is on a mission to build innovative character editing tools to help global storytellers tell their best stories.
at IBC, we demonstrated that visual dubbing at the highest quality is our first frontier towards that mission.
Vishal retweeted
not a demo. not a clip. a full feature film, shot in Telugu, lip-synced into Malayalam, now streaming on Netflix.
powered by sync. labs.
Vishal retweeted
Today, @JudgmentLabs opens for general access.
We believe long running agent systems will generate more exhaust data than the world can currently process.
Hidden within agent traces are rich records of how to make these systems better: what users ask for, when agents stumble, and how agents interact with the external world.
Judgment is the most performant monitoring and improvement layer for long-horizon agents.
We’re excited to see what you build!
Vishal retweeted
try to not use agents to read papers or blogs in the first pass until youve given it a read, the idea is to make your brain process the information with your current cognitive level and let it think and get inspired for ideas and doubts. when you feel like youve done enough homework, then switch to asking agents to explain and then pull up references and supplementary materials to further solidify. dont let the agent skew your thinking to what it thinks.
with agents, you need to feel smarter and not let agents do all the smart work for you. dont be worried about speed and be worried about deep understanding.
Vishal retweeted
for the first time in a long time i can feel my mania coming back. in the best way possible. my hunger for life. excitement for building things. vision for what they could become. i feel myself regaining the child like wonder i had lost for so so long
life. what a lovely thing..
Vishal retweeted
can we trust decision models for evals?
@HamelHusain and I tested top decision models like Jev, d1-ultra, clef, pplx-decider, openai decisions, etc. on our dataset
we explored failure detection, probability calibration, threshold optimization, and the practical tradeoffs of using them for evals.
give it a read: hamel.dev/blog/posts/decisio…
Vishal retweeted
you already posted it. now make it say something new.
sync. now allows you to import a link from any social media platform and turn your existing content into your next post. also with our api.
good content deserves another go.
at this point i think IDEs are bloat, since almost all of us are using agents to write the code, we don't need IDEs to look at the files.
we can use neovim or vanilla vim if we want to browse through the changes (most people aren’t even reading the diffs anymore), just raise a pr and read your changes there, comment on it, and have the agent fix it
