Skip to content
Trending storyWatching

DeepSeek open-sources DSec: elastic sandbox infrastructure for agentic training

1 report1 sourceupdated 3 days ago

What happened

Summary

DeepSeek 发了一篇论文,把他们内部用来大规模训练 agent 的沙箱系统 DSec 公开了。核心思路是让模型在隔离的沙箱里反复练习使用真实工具,而不是靠模拟数据。系统能同时跑 10 万个沙箱,启动一个沙箱只要 11 秒,单个成本大概 3 美元。论文里讲了怎么用快照、缓存和按需分配把延迟和开销压下来,也给了训练 agent 时的效果数据。目前只放...

Coverage

Follow the reports to see the story from different sides.

Sep 27
  1. Hacker News front pagePick
    DeepSeek open-sources DSec: elastic sandbox infrastructure for agentic training

    DeepSeek published a paper on DSec, their internal sandbox system for training agents at scale. The idea is to let models practice with real tools in isolated environments that scale elastically. It handles 100K concurrent sandboxes, 11-second startup latency, and roughly $3 per sandbox. The post doesn't mention a code repo—only the arXiv paper is available so far.

Heat over time

Not enough continuous observations to draw a trend yet.