When AI agents no longer require you to keep your computer screen awake and on standby, but can operate autonomously after you turn off the lights and leave, the boundaries of workplace collaboration are being fundamentally reshaped.<\/p>\n
U.S. artificial intelligence company Anthropic announced on Tuesday that its AI agent Claude Cowork, designed to execute digital tasks, has officially broken free from desktop constraints, expanding its functionality to mobile devices and web browsers. This means users no longer need to keep their laptops constantly awake as before; they can now let the agent work continuously in the background, even overnight, handling scheduled tasks.<\/p>\n
This move marks Silicon Valley’s battle over AI agents officially entering a new phase of “always-online” capability.<\/p>\n
From “Can’t Turn Off the Computer” to “Burning the Midnight Oil”<\/p>\n
Previously, although Claude Cowork demonstrated impressive task-handling abilities upon its release earlier this year\u2014such as organizing cluttered screenshot files or scheduling appointments\u2014its operation suffered from a frustrating flaw: the computer had to remain awake, and the application interface had to be open. To avoid interrupting the agent, many early users even developed the habit of “half-closing” their laptops.<\/p>\n
Anthropic’s latest update completely resolves this pain point. In a demo video released by the company, a user asks Cowork to prepare materials for a business renewal meeting the next day. With a single command, Cowork not only needs to pull data from past emails, Slack channels, meeting notes, and recent online discussions, but also generate reference documents and draft emails. Previously, all of this had to be completed while the computer was active; now, even after the user clocks out and rests, the agent can continue receiving late-night messages and processing tasks.<\/p>\n
Wired magazine, after an early hands-on experience, commented that while Claude Cowork is still not perfect and faces security risks such as injection attacks, its impact on how ordinary users interact with their devices is a “leap forward.”<\/p>\n
The “Agent Storm” from OpenAI and Google<\/p>\n
Anthropic’s update is not an isolated case but a microcosm of Silicon Valley’s full-scale shift toward “semi-autonomous, always-resident” agents. This trend was ignited by the geek-built agent OpenClaw, which exploded in popularity in early 2026, with early adopters even running it 24\/7 to take over their online lives.<\/p>\n
Tech giants quickly followed suit. OpenAI not only hired OpenClaw’s creator and launched the adaptive agent Codex, but also rolled out Codex Remote with similar functionality in June this year, allowing users to control desktop agents via smartphones. In July, OpenAI further updated its iOS app to support creating, searching, and managing Codex tasks directly within conversations.<\/p>\n
Meanwhile, an internal report recently released by OpenAI reveals the astonishing speed at which agents are proliferating in the workplace. The report notes that AI agents are redefining the fundamental unit of knowledge work\u2014shifting from single interactions to tasks that can be delegated and autonomously executed over extended periods.<\/p>\n
Data shows that by 2026, Codex accounted for 99.8% of OpenAI’s total weekly internal token output, becoming the primary AI tool for every department, including legal, finance, and recruiting. A more critical shift lies in the leap in task depth: as of May 2026, 80.6% of users had submitted tasks to Codex requiring more than 30 minutes of human work, 70.2% submitted tasks exceeding one hour, and tasks corresponding to over eight hours of work showed the fastest growth rate.<\/p>\n
“User habits are quietly changing: from short, quick interactions to more complex, longer-cycle task delegation,” OpenAI summarized in the report.<\/p>\n
From Office to Lab: AI Agents Enter Scientific Research<\/p>\n
Beyond the workplace, AI agents are becoming scientists’ “best collaborators.” Google DeepMind recently published its latest Co-Scientist system in the journal Nature\u2014a multi-agent collaborative system built on Gemini.<\/p>\n
Unlike handling office tasks, Co-Scientist attempts to solve the “needle in a haystack” pain point in scientific research: how to find the right scientific hypotheses amidst vast amounts of information. The system simulates the complete cycle of scientific thinking, with specialized agents responsible for generating hypotheses, critical review, and iterative evolution.<\/p>\n
Most striking is its unique validation mechanism. The system draws on the competitive mechanisms of AlphaGo and AlphaStar, having agents engage in scientific debates. Through an “idea tournament” style of pairwise comparison and simulated debate, the system continuously filters and eliminates, while deeply cross-referencing scientific literature and professional databases to ensure hypotheses are logically rigorous and evidence-based.<\/p>\n
According to Google DeepMind, Co-Scientist has already produced tangible impact across multiple fields: one team used it to discover a previously overlooked drug for treating liver fibrosis; a research group compressed the analysis of a massive cellular senescence screening dataset from months to days; and an aging biology company generated new hypotheses based on it that were validated in the lab. Participating scientists stated that the AI agent’s most important contribution lies in improving efficiency and dramatically shortening the cycle for achieving scientific breakthroughs.<\/p>\n
Reshaping the Boundaries of Collaboration<\/p>\n
Whether it’s OpenAI’s emphasis on “expanding the radius of work” or Google DeepMind’s positioning as a “research partner, not a replacement,” these cutting-edge explorations all point to the same core proposition: when agents can take on tasks that previously required specialized skills, human workflows, functional boundaries, and skill values will all face reassessment.<\/p>\n
The Claude Cowork version released by Anthropic will first roll out in beta to subscribers of the $100 per month Max plan, subsequently trickling down to the $20 per month Pro plan. A report released concurrently by the company shows that white-collar workers are increasingly integrating its tools into their workflows, with “business processes and operations” and “content creation and copywriting” being the two categories with the highest recent usage.<\/p>\n
Silicon Valley’s giants are clearly not content to let agent tools serve only geek developers. Their common strategy is to stop launching standalone new applications and instead embed these capabilities directly into the chatbots already present on millions of users’ phones. When AI agents truly achieve ubiquitous “background operation,” what humans need to contemplate, as industry observers suggest, is perhaps this: the real question is where people should place their time and energy.<\/p>\n","protected":false},"excerpt":{"rendered":"When AI agents no longer require you to keep your computer screen awake and on standby, but can…\n","protected":false},"author":2,"featured_media":99253,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[9],"tags":[8652,51766,53,5914,27796,2317,5044,2408,132,7543,866,157],"class_list":["post-99252","post","type-post","status-publish","format-standard","has-post-thumbnail","category-google","tag-alphago","tag-alphastar","tag-anthropic","tag-claude-cowork","tag-co-scientist","tag-codex","tag-deepmind","tag-gemini","tag-google","tag-google-deepmind","tag-nature","tag-openai"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/99252","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=99252"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/99252\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/99253"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=99252"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=99252"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=99252"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}