Re 5/ V1: Unifying Generation and Self-Verification for Parallel Reasoners Paper: https://arxiv.org/abs/2603.04304
Re 5/ V1: Unifying Generation and Self-Verification for Parallel ReasonersPaper: https://arxiv.org/abs/2603.04304
Re 5/ V1: Unifying Generation and Self-Verification for Parallel ReasonersPaper: https://arxiv.org/abs/2603.04304
Re 2/ ThunderAgent: A Simple, Fast and Program-Aware Agentic Inference SystemPaper: https://arxiv.org/abs/2602.13692
Re 4/ Escaping the Verifier: Learning to Reason via Demonstrations (RARO)Paper: https://arxiv.org/abs/2511.21667
Re Read the blog: https://www.together.ai/blog/icml-2026See us at ICML: https://www.together.ai/icml-2026
Re 1/ DSGym: A Holistic Framework for Evaluating and Training Data Science AgentsPaper: https://arxiv.org/abs/2601.16344
Our research team has 9 papers at ICML next week! Spanning the full stack from frontier agents to GPU kernels, we're excited to share what our…
good days when you could just wait for a faster Nvidia gaming GPU and it'd eventually fall into your lap. don't try that that stuff nowEric…
Fable is very good, and having it taken away after getting early access delayed a few things I was working on and made me defer others…
I mean, could be worse. At least you're not dying for the glory of conquering (maybe) (temporarily) a bumfuck nowhere village called like "Malaya Dickensovka", after…
Claude Fable 5 will be available again globally tomorrow.After a series of productive conversations with the US government, we're redeploying the model with a new set…
RT Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)Re "All Chinese actors" is barely a meaningful category, and the US AI (whether open or closed) is heavily Chinese…
Whale Camp is winning… hearts and minds at leastwill brown: i’ve come around to understanding that the golden path application of OPD is MOPD to enable…
«Horizon is a line you can never reach»And true agents will be able to keep goingCunxiang Wang: From Long-horizon Tasks to Endless Frontier Tasks.The next frontier…
Trump is an insecure personalist despot and Dario is an insufferable libshit egghead from Princeton, with a funny name tooTom Brown got it sorted out like…
RT ZainWe gave a 2 hr deepdive on how to build inference engines that handle trillion token agentic workloads at @aiDotEngineer.Will drop slides and detailed walkthrough!
RT 🌅 сатья-юга в моей квартиреI hate this attitude so much.We've basically discovered the way to make humanity smarter, healthier and better forever with no downsides…
RT Matthew AtkinsonOne unexpectedly great detail from @aiDotEngineer: the daily conference newspaper. Nice to flip through something with a hard stop, no algorithm trying to turn…
RT Lei LiOur MOPD from MiMo-V2-Flash has been widely adopted in modern post-training pipelines. Now the paper is out with more details & comparison.Check it out:…
weirdly specificAndrew Curran: I'm posting this prediction now so I can quote it later. There has been a significant breakthrough in architecture - specifically around memory…
RT Crooked MediaShare this so Elon Musk can’t pretend kids didn’t die when he destroyed USAID.
RT Peter Steinberger 🦞Apparently we didn't talk enough about w̶o̶r̶k̶f̶l̶o̶w̶s̶ loops yet! See ya there!Warp: @steipete is now joining us for Crafting Software Factories!📅 6pm Wed…
RT rachael de foe is in sf ✈️what do YOU do while waiting for ai to cook? 🍳🧑🍳: @WilliamBryk @vincent_koc @altryne #paulinebrunet @swyx @0thernet @vincent_koc @charles_irl…
Surprisingly good and thoughtful set of predictionsoccasional hopium notwithstanding, I agreebayes: http://x.com/i/article/2072056179295965184
RT John Schulmanhiring post-training hackers to make Tinker even better!Tinker: We're looking for Tinker fanatics to help the world fine-tune models. Join us!https://job-boards.greenhouse.io/thinkingmachines/jobs/5202369008