Onomos @o is for autonomous Get in touch

Writing

DeepSeek’s super-agent is a signal. Start now.

Autonomous systems, where software agents do the work and people make the decisions, are the way to go, and DeepSeek’s super-agent is a signal that CEOs should get ready. We know how to implement them and fit them to your organization. The sooner you start, the better for your business results.

Book a call aut@onomos.ai

Onomos

The signal

On 1 November 2026 the team behind DeepSeek Harness, DeepSeek’s agent software, published “Exploring Super-agent Design”. By the next morning it had about 1.8 million views. The idea: one agent, meaning software that carries out work on its own, takes requests and hands out the work, others do it, and people move up to decisions and sign-off. Amoeba Farm’s short animation shows it in under a minute.

Animation by Amoeba Farm, shown from its post on X. Watch it on X. Any test figures in the post come from small simulated runs that DeepSeek itself calls indicative only.

It is strong work, and DeepSeek is clear about its status: the article discusses a design, not existing features, and its prototype is research code, not an official product. For a CEO, the product is not the news. The direction is. AI models need less and less help on any single task, so the limit on getting more done is moving to human attention. Our reading: how a company assigns, owns and checks work now decides how much it can hand over. Autonomous systems are the way to go, and companies should get ready now.

We build autonomous systems

Our work does not come from DeepSeek’s article. Onomos and agent.ceo belong to the same company group, and agent.ceo was running before that article appeared. It is an autonomous development system: agents in organizational roles plan, build, test, review and ship the platform they work on, and people keep four decisions: credentials, money, accepting security risk and strategy. We keep improving it.

We use many methods, some developed by us and some not, and more will come. We have also worked with enterprises on this: for a European enterprise software vendor with more than a thousand employees, an audit of how it builds its agents, and a software development process run by agents; for a global technology group listed in Amsterdam, preparing a platform of autonomous agents for its developers. Both recognized the need to prepare for autonomy.

Fitted to your organization

Every company has its own way of working: decisions made in meetings, exceptions known to one person. So we start from your company, not from a pattern, and we consult on the path, not on a product.

Tasks move through four stages over time, each at its own pace, and not every task needs to reach the last.

  1. Human-ledA person reviews every part. New ideas start here.
  2. AgenticwareAgents write the work and its checks, with or without a person checking.
  3. DelegateAgents keep improving the work, version by version.
  4. Agentic unitAgents own the outcome.
The same task moves through the stages over time. Different tasks sit at different stages at once.

Start now

Start with one process. In three weeks, with its owner, we sort every step by stage, set the measures and what counts as done, and leave a plan for a first step you can fund, with owners, effort and dates. You keep everything delivered. If you then want a leader for the change, we stay on, eight weeks at a time.

Getting a company ready takes time. The sooner you start, the better for your business results. Book a call or write to aut@onomos.ai.

Sources

  1. DeepSeek Harness (2026). “Exploring Super-agent Design: Towards a More Autonomous Agent System.” X, 1 November 2026. Machine translation of a Chinese original. View count as shown on 2 November. https://x.com/DeepSeekHarness/status/2116857257245987236
  2. Amoeba Farm (2026). Post and animation on the DeepSeek write-up. X, 1 November 2026. https://x.com/AmoebaFarm/status/2116955940402377102
  3. DeepSeek. dsh-superagent-prototype: research prototype, not an official DeepSeek product. MIT license. https://github.com/deepseek-ai/dsh-superagent-prototype
  4. agent.ceo (2026). “Beyond agents that do tasks: building organizations that improve how they work.” 15 October 2026. https://agent.ceo/blog/agentroots-organized-agentic-development