[ByteDance OpenViking Core Contributor Open-Sources LoopX: Enables Agents to Run Continuously for 200 Hours Without Memory Loss or Deviation] According to monitoring by Beating, ByteDance AML Senior Machine Learning Engineer and OpenViking core contributor Huang Ruiteng has open-sourced LoopX. It is a control system designed for long-term agents, allowing agents like Codex and Claude Code to continue working toward their original goals even after tasks span multiple days and interruptions. LoopX has publicly shared two real task trajectories spanning 220.7 and 272.9 hours. During this time, the agents underwent multiple rounds of execution, waiting, human judgment, model switching, and task recovery, yet they were still able to retrieve the current goal, existing evidence, and next steps. LoopX places goals, to-dos, permissions, evidence, and waiting conditions outside the model's context. Agents take only small steps at a time, verify results, and then write back the updated state. Whether switching conversations, changing models, or restarting programs, they can continue from the latest progress. OpenViking, which Huang Ruiteng contributed to developing, is responsible for storing and retrieving agents' memories, data, and skills, while LoopX manages task progress, next steps, and when human judgment is required. The former acts like long-term memory, while the latter functions more like a project manager and actionable kanban board. Currently, LoopX has been applied to automatic code issue repair, AutoML experiments, and long-term research. [Original Link]