🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Brb, co-authoring a paper with C. Opus: Turning Up the Heat Again: A coherence study of Min-p

Evidence is cool, but have you looked at the data? The devil is in the detail... especially with LLM generations

Feels like every major AI lab has a DeepResearch model, yet the publications on how to get a performant model are non-existant

Why do people hate Alexander Wang so much? I don’t get it, you mad he got a bag from Zuck?

Respect to Ilya for believing in the SSI mission and not being a sellout. He is more concerned with superintelligence than money

veRL added support for tool calling agents a few months ago, but now they are exceedingly streamlining it in clean loops with ultimate control in the user's hands👀

these types of videos used to generate a lot of attention on X, but the reign is over. figuring out the new algorithm is like playing jenga

short videos and <5 word posts now do well. grok got brainrot from watching tiktok and now it's spilling over

I have gobbled up bookmarks on new RL trainers for multi-turn RL, here are the noteworthy mentions: • RL2: Ray Less Reinforcement Learning. The NanoGPT of RL with it's small and hackable size (<1k lines) • AReaL (Ant Reasoning RL): LLM generation runs in a streaming manner,

that's it, that's the post

OpenAI employees shoveling money into their bank accounts so fast that they don’t even notice charges sub $1000

OpenAI’s open-source model is dropping after GPT-5 and it’s going to be o3-mini level. Sam Altman personally whispered this to me in a dream