OS World

Agents at Scale: Inside MiniMax's Model and the Infrastructure Behind It — Dan Fu and Olive Song

Jul 31, 2026 · 20:14

Olive Song, RL lead at MiniMax, and Dan, Together AI's VP of Kernels, argue that open-weight MiniMax M3 is closing the gap to frontier labs via community feedback and inference optimization. Dan details the day-zero stack: GPU kernels for M3's sparse attention, KV cache at million-token contexts, and adapting to agentic workloads that upload codebases. Song says M3 trains multimodal from scratch to prevent text-vision collapse, and uses RL with environment design and rewards for long-horizon tasks like replicating a 12-hour iClear paper, plus validation/test splits to avoid hacking. Together's Parallel Kernel Bench holds unsolved problems, so overfitting it yields kernels the company deploys. Both see self-evolution and better GPU utilization as why open models keep accelerating.

How to Improve Your Agents: Academic Lit Review

Feb 22, 2025 · 39:02

Joe from Columbia University and founder of Arklex AI explains research on AI agents, focusing on improving their reasoning and planning through self-reflection, test-time compute, and tree search methods, without relying on human supervision. He details the TriPath method, which uses larger models to edit smaller models' feedback for better self-improvement, achieving up to 48% accuracy on math benchmarks. He shows how multi-color tree search (MCTS) with contrastive reflection and multi-agent debate (RMCTS) outperforms other search methods on Visual Web Arena and OS World, achieving top non-trained results. He proposes exploratory learning, where models learn from search trajectories rather than optimal actions, improving performance under compute budgets. Finally, he presents the Arklex open-source agent framework, combining machine learning, systems, and security for practical multi-agent orchestration.