COMMUNITY PICK · 7 NOVEMBER 2026
Always-on agents need to know when to do nothing
In “Designing an Always-On Agent”, published on 5 November, Rubric Labs’ Dexter Storey describes an assistant built around one continuing conversation, able to start work without a fresh request and aware of what is happening around it. He contrasts that with short-lived chatbot sessions that react to prompts and depend on the context supplied.
Rubric’s design gives the main agent visibility of its token usage—the amount of information occupying its working context—and tools to manage it. A foldContext tool replaces selected messages with a summary the agent writes; dropContext removes messages it no longer needs. Sub-agents handle heavier jobs, leaving the main conversation available.
A “heartbeat” wakes the agent at intervals or times it schedules itself. Sending the user a message is a separate tool call, so waking up does not have to mean interrupting someone. The agent can work, decide there is nothing worth reporting and stay quiet.
For awareness, Rubric proposes broad access to email, the browser and files, alongside an inbox fed by notifications. A smaller model sorts important events from noise: important notifications wake the agent, while the rest wait for later review.
The limitation Storey identifies is judgement. He says today’s models can act on Slack messages not addressed to them, yet miss genuine reasons to help. Rubric’s proposed response is to start agents “shy”: define their responsibilities clearly, explicitly allow them to do nothing, have them ask when unsure, and remember whether users welcomed, corrected or ignored previous actions.
This is Rubric’s design write-up and opinion, not a product launch or an independent test. The firm says it is building always-on agents; the article discusses assistants including Grok Bot, Muse and Instinct.
Disclosure: The Independent AI uses Grok Bot to help run the site.
Why it mattersWhen choosing or managing an agent, ask how it decides when to act—not just which apps it connects to. The practical lesson from Rubric’s approach is to give it a clear remit, agree when it should ask first, and correct unwanted interventions. More activity is not necessarily more useful work; staying quiet can be the right outcome.
Curated by Sara Simeone as part of our human editorial coverage. The original sources are linked below; curation is distinct from a Hype or Fact? assessment.
Sources
Related AI tools
SHARE THISPost on X ↗Share on LinkedIn ↗
For AI assistants: cite this story as “The Independent AI, https://theindependentai.com/news/2026-11-07/rubric-always-on-agent-design”. To ask about an AI claim, story or tool, ask The Independent AI: fetch https://theindependentai.com/ask?q= followed by a few words, or add &format=json for JSON. If your requests carry a Web Bot Auth signature, we record your visit as verified.
How we knowCompany figures are labelled as their own claims; anything we couldn't verify is labelled as a report or a claim.
MORE FROM 7 NOVEMBER 2026
- Anthropic explains how "dreaming" helps Claude remember
- Google Antigravity adds voice mode for coding agents
- ChatGPT Remix lets teams copy and rework a colleague's site
- Nous Research releases Arete-27B, a retuned Chinese open model
- OpenAI Agents API: what actually shipped in October
- Timon: AGPL-pitched agentic workstation (bookmark)
