Pinned post
I've written about 500+ machine learning papers in the past year. Here are some of my most popular threads: [1/n]
Davis Blalock
Verified account This account is verified. Learn more@davisblalock
Research scientist @GoogleDeepMind. Past: @Databricks, first hire @MosaicML, @MIT PhD. I post about AI papers + sometimes the business side.
171 Following 15K Followers 1.4K Posts
Latest posts 5
A 10-person team with $4M just trained a better speech-to-speech model than the incumbents.
I wouldn't have thought this was possible, but it appears software is now more skill- and caring-limited than manpower-limited. (And Moin et al. are extremely skilled and care a lot).
I wouldn't have thought this was possible, but it appears software is now more skill- and caring-limited than manpower-limited. (And Moin et al. are extremely skilled and care a lot).
Introducing Phonic-1, the most advanced speech-to-speech model for agentic tasks — from @phonic_co, the speech-to-speech model company.
We’ve had a breakout year and raised a $27.5M Series A led by @firstround with participation from @Lux_Capital, @felicis, and Definition.View quoted post on X
My friend's new company is offering way faster inference for the big open-weight models. Pretty surprised there was even this much room to beat the incumbents, let alone that they did it with their stack so nascent. Bodes well.
Inco AI is entering public beta with the ⚡️ fastest ⚡️ #inference endpoints for #Kimi K3, #MiniMax M3, #GLM 5.3, and #GLM 5.3 Flash.View quoted post on X
Ranking Number 1 on the respective @ArtificialAnlys provider leaderboard for output speed.
Read more: inco.ai/blog/inco-plat…
🚀 Today we’re releasing FlashOptim: better implementations of Adam, SGD, etc, that compute the same updates but save tons of memory. You can use it right now via `pip install flashoptim`. 🚀
arxiv.org/abs/2602.23349
A bunch of cool ideas make this possible: [1/n]
arxiv.org/abs/2602.23349
A bunch of cool ideas make this possible: [1/n]
