Nearly half a year of silence. We spent it studying one problem: how far RL can scale.
MiMo-V2.6 is in the middle of its RL run right now. Three things we scaled: compute (~2B tokens per step, 1568 prompts × 16 rollouts, fully async), environments and harnesses (multi-task agentic RL, mixed across multiple harnesses in one run), and grader compute (agentic in-group credit assignment, with test-case and rubric-based rewards). We'll open-source the details piece by piece over the coming weeks.
Streaming the run: https://t.co/ZSxahzJRju
Open-weight AI models aren't just a technical preference — they're a safeguard. The real risk of Superintelligence isn't misalignment, it's who it's aligned to. Silicon Valley, Washington, or Beijing shouldn't get a monopoly on the most powerful technology in human history. Open weights = checks and balances for the AI age.
Introducing OpenMythos
An open-source, first-principles theoretical reconstruction of Claude Mythos, implemented in PyTorch.
The architecture instantiates a looped transformer with a Mixture-of-Experts (MoE) routing mechanism, enabling iterative depth via weight sharing and conditional computation across experts.
My implementation explores the hypothesis that recursive application of a fixed parameterized block, coupled with sparse expert activation, can yield improved efficiency–performance tradeoffs and emergent multi-step reasoning.
Learn more ⬇️🧵
@AravSrinivas Hi Aravind, many of our Pro subscribers are seeing their subscription disappeared today while still being charged as usual. Your support team is not responding. Very disappointing.
• 科斯定理(Ronald Coase, "The Nature of the Firm", 1937):这是解释“公司”为何存在的终极理论。科斯认为,公司的存在是为了降低交易成本(沟通、信任、缔约的成本)。如果完全在市场上找人合作成本太高,人们就会成立公司,用内部的管理指令来代替外部的市场交易。
• 推论:当 AI Agent 极其聪明、可靠且成本极低时,它能以光速完成需求匹配、谈判、履约(甚至结合智能合约)。这意味着外部市场交易成本将无限趋近于零。当外部交易成本 < 内部管理成本时,庞大的科层制公司就失去了存在的经济学依据,组织规模将不可避免地向单人+多 Agent 的形态收敛。