Understand this story
The DeepSeek Harness team published "Super-agent Design Exploration: Toward a More Autonomous Agent System" on its official WeChat account, proposing a five-module approach—dual-layer architecture, event-driven design, asynchronous multi-channel interaction, storage-compute separation, and an external Watchdog—to support an always-online, concurrently responsive Agent. The team said an internal Super-agent instance built on this approach is already running continuously, participating in issue reproduction and fixes, code review responses, and DeepSeek Harness's own iteration. In preliminary ablation experiments, the Super-agent's task completion time p50 and p95 were 12.4 minutes and 15.7 minutes, respectively, outperforming a single-Session solution with a Subagent tool. A same-day report by QbitAI further explained that the upper-layer Router is responsible for receiving messages, assigning tasks, and maintaining global context, while the lower-layer Worker executes specific tasks in parallel; in a simulated environment with about 300 messages and 20 work requests, repeated tests 5 times showed that the Super-agent's median task completion time was 12.4 minutes and its 95th percentile was 15.7 minutes, while the traditional single-Session solution's 95th percentile under a neutral prompt was 29.5 minutes.
AI-generated from public reports
Latest development · Report
DeepSeek proposes Super-agent architecture: a two-tier design with Router and WorkerDevelopments
1 confirmed developments
DeepSeek Harness team shares exploration of Super-agent design
Source:East Money
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 1
IT Home · Artificial Intelligence
DeepSeek proposes Super-agent architecture: a two-tier design with Router and Worker
The official DeepSeek Harness team released "Super-agent Design Exploration," proposing a two-tier architecture composed of a Router layer and a Worker layer. The Router handles human interaction, maintains global context, and dispatches tasks, while Workers execute concurrently in isolated sessions. The team says the architecture has four characteristics—continuous, efficient, proactive, and persistent—and incorporates mechanisms such as event-driven operation, Inline & Spill, chain recall, speculative execution, and Watchdog. In preliminary experiments, the p50 and p95 task completion times for Super-agent were 12.4 minutes and 15.7 minutes, respectively, while the p95 for a single-session approach under neutral prompts was 29.5 minutes; the team notes that the experiments were based on synthetic workloads and a simplified environment. The article also provides a prototype implementation named dsh-superagent-prototype and makes clear that it is not an official DeepSeek product.
IT Home · AIRelated discussion
IT Home · AI joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Zhidx
DeepSeek Harness unveils Super-agent architecture, using five major modules to take over session scheduling
The DeepSeek Harness team has unveiled the underlying architecture of Super-agent, an autonomous Agent exploration project, bringing classic computer systems ideas such as scheduling, caching, compute-storage separation, and Watchdog into Agent design. The architecture consists of five modules: a dual-layer architecture, event-driven design, asynchronous multi-channel interaction, compute-storage separation, and an external Watchdog. Its goal is to enable the Agent to stay continuously online, advance multiple tasks in parallel, and proactively respond to environmental events. In preliminary experiments, Super-agent's p50 and p95 task completion times were 12.4 minutes and 15.7 minutes, respectively. After removing subscription passthrough, the number of Router decisions increased by about 30%, and read-cache tokens rose from 23.23 million to 31.77 million. The team said the experiments were based on synthetic workloads and a simplified environment, and that this plugin is only a prototype for validating the technical approach, not an official DeepSeek product.
Oct 31
East Money
The DeepSeek Harness team publishes a technical article on Super-agent design exploration
The DeepSeek Harness team published "Super-agent Design Exploration: Toward a More Autonomous Agent System" on its official WeChat account, proposing five modules—a dual-layer architecture, event-driven design, asynchronous multi-channel interaction, storage-compute separation, and an external Watchdog—to support an Agent that is continuously online and responds concurrently. The team says an internal Super-agent instance built on this approach is already running continuously, participating in issue reproduction and fixes, code review responses, and iteration of DeepSeek Harness itself. In preliminary ablation experiments, the Super-agent's task completion time p50 and p95 were 12.4 minutes and 15.7 minutes, respectively, outperforming the single-Session approach with Subagent tools.
East MoneyRelated discussion
East Money joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Yicai
DeepSeek Releases Super-agent Architecture; Internal Instance Already Participates in Code Review and Fixes
The DeepSeek Harness team published a technical article titled "Super-agent Design Exploration: Toward a More Autonomous Agent System," proposing a dual-layer architecture, event-driven design, asynchronous multi-channel interaction, storage-compute separation, and an external Watchdog to break through the scale limits of multitasking and team collaboration. The team said that an internal Super-agent instance built on this approach is already running continuously, routinely participating in issue reproduction and fixes, code review responses, and DeepSeek Harness's own iteration; when members report software issues in group chats, even if no one @s the AI, it can identify potential defects and attempt in the background to reproduce them and find fixes.
YicaiRelated discussion
Yicai joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Shao Meng
The DeepSeek Harness team shares the Super-agent architecture design: using...
The DeepSeek Harness team shares its Super-agent design exploration, arguing that single-task Agents are already strong enough and that the bottleneck has become the roughly 3–5 active Sessions a person can monitor at once; the Agent is therefore made into an entity that is continuous across channels, parallelizable, event-driven, and persistently online in the cloud. The proposal includes a two-layer architecture of a resident Router plus Workers with isolated contexts, two-level Trigger filtering with direct Thread pass-through, asynchronous ReAct, separation of the cloud Loop from the device-side daemon, and an external Watchdog. In synthetic load tests, task completion time p50/p95 was about 12.4/15.7 minutes, better than the roughly 29.5-minute p95 of a single Session plus Subagent, at the cost of higher read-cache tokens.
QbitAI
DeepSeek Unveils Super-agent Architecture: Two-Tier Router and Worker Design
DeepSeek has published an engineering-oriented architecture article titled "Super-agent: Toward More Autonomous Agent Systems," proposing that an upper-layer Router is responsible for receiving messages, assigning tasks, and maintaining global context, while lower-layer Workers execute specific tasks in parallel. In a simulated environment with approximately 300 messages and 20 work requests, repeated tests 5 times showed that the Super-agent's median task completion time was 12.4 minutes and its 95th percentile was 15.7 minutes, whereas the traditional single-Session approach had a 95th percentile of 29.5 minutes under neutral prompts. The article also introduces two-level event filtering, an asynchronous ReAct Loop (already implemented in DeepSeek Harness but not enabled by default), separation of computation and state from the execution environment, and a Watchdog supervision mechanism.
Heat trend for this story
- Current
- 51
- Period peak
- 54 Nov 1 03:00
- 24-hour change
- —
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.