Oriel — AI & technology worth noticing, every day.

CoworkOS:Scaling Agent Coworkers for Real-world Automation

Paper · https://arxiv.org/abs/2611.03178

Rapid advances in cowork agents are shifting human-AI collaboration toward real-world automation, delivering software, designs, and other artifacts. However, such productivity work demands both high-quality delivery on complex productivity tasks and continued monitoring and decision-making over long horizons, a combination that remains an open challenge. We introduce CoworkOS, a general-purpose harness for long-horizon work that separates task strategy from execution management: the model decides actions, delegation, and acceptance, while the runtime maintains state and feedback. Its three-layer architecture enables direct execution or delegation to Subagents, synchronously or asynchronously, under stable responsibilities. Through a shared application layer, humans and agents inspect and revise the same evolving documents, presentations, CAD models, and Blender scenes, so human-in-the-loop participation can enter both through conversation and through direct modification of the work itself. To sustain continuity across agents and sessions, CoworkOS further introduces hierarchical workspaces and persistent semantic memory, so working files, key decisions, and long-running Job logs remain shareable, reusable, and resumable after interruption. Across 12 benchmarks and 6 categories of real-world applications, CoworkOS proves highly competitive with frontier agent applications such as Codex and DeepSeek Harness. As a sustained-work case study, CoworkOS uses GLM-5.3 as its backbone to conduct RSI-driven automated post-training on Qwen3.6-35B-A3B. Over five rounds, the model's average score across eight benchmarks rose from 33.19 to 42.95 (+9.76). Looking ahead, we believe CoworkOS opens a meaningful path toward general cowork-agent intelligence.

First seen by Oriel on .

AI summary

CoworkOS is a general-purpose runtime for long-horizon agent work that separates model decisions from execution management and supports human collaboration on evolving artifacts.

  • Its three-layer architecture supports direct execution or synchronous and asynchronous delegation to subagents with stable responsibilities.
  • Humans and agents can inspect and revise shared documents, presentations, CAD models, and Blender scenes through conversation or direct edits.
  • Hierarchical workspaces and persistent semantic memory preserve working files, decisions, and job logs across agents, sessions, and interruptions.
  • Across 12 benchmarks and six application categories, it was reported as highly competitive with Codex and DeepSeek Harness.
  • In a post-training case study using GLM-5.3 on Qwen3.6-35B-A3B, average scores across eight benchmarks increased from 33.19 to 42.95 over five rounds.

#ai agents #human-ai collaboration #persistent memory #model post-training

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hugging Face Papers: 100, latest observation ; +14 over observed period since

Comments

Comment count reported by the source.

  • Hugging Face Papers: 4, latest observation ; +2 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Papers2611.03178https://arxiv.org/abs/2611.03178 Read discussion