Detailed Analysis
Anthropic has expanded Claude's "Cowork" feature to run scheduled tasks asynchronously in the cloud, meaning automated workflows now execute independent of whether a user's local machine is powered on. The example cited—an automated 6am "client prep" routine that reviews message threads and transcripts, compiles a brief, and drafts (but does not send) a follow-up—illustrates Anthropic's push toward persistent, background AI agents rather than session-bound chatbots. When the automation encounters a decision requiring human judgment, it escalates the query directly to the user's phone, effectively creating a desktop-to-mobile handoff loop for ongoing work. This cloud-based execution model, seemingly running on Anthropic-managed containers based on user speculation in the replies, represents a meaningful architectural shift: Claude is no longer just a tool invoked in a browser tab but a semi-autonomous agent that persists and acts on a schedule.
The public reaction captured in the replies reveals a community sharply divided on execution, even among people generally receptive to agentic AI. One user proposes a genuinely useful workflow pattern—having Claude leave a "where we left off + next step" note in a working folder at the end of a desktop session so a later mobile pickup has full context—which points to a real design opportunity around continuity of state across devices and sessions. This kind of user-generated best practice often signals where a company's own roadmap should be heading, since it shows organic demand for session memory and handoff mechanisms that reduce redundant context-setting, a persistent pain point in multi-turn, multi-device AI usage.
Other replies are more critical and reflect broader anxieties in the Claude user base about product direction, pricing, and reliability. Complaints about the UI merging "Chat" and "Cowork" into a single interface—rather than keeping distinct surfaces for conversational chat, delegated background work, and coding (Claude Code)—echo a common tension in AI product design between consolidation for simplicity and specialization for power users. Similarly, confusion over usage limits ("we don't have Cowork specific limits, so what does doubling the usage limit for Cowork actually mean") suggests Anthropic's tiered pricing and quota system remains opaque to subscribers, a recurring friction point as the company tries to monetize increasingly compute-intensive agentic features. The comparison to GLM (a cheaper, more permissive open-weight model alternative) and accusations of over-restriction on a $200/month Max plan tier highlight the competitive pressure Anthropic faces from lower-cost or open models that impose fewer safety-driven refusals, a tradeoff central to Anthropic's brand identity as a safety-focused lab.
More broadly, this development sits within the industry-wide shift toward "agentic" AI—systems that don't just respond to prompts but operate continuously, make decisions autonomously within bounded parameters, and only surface judgment calls to humans when necessary. Competitors like OpenAI and Google have been building similar scheduled-task and background-agent capabilities, so Anthropic's Cowork expansion is best understood as part of an industry race to own the "AI as autonomous coworker" positioning rather than "AI as chatbot." The hostile and skeptical replies in this thread—ranging from pricing complaints to inflammatory rhetoric about investor honesty—also underscore that as these companies push toward more autonomous, higher-stakes AI products, they face mounting scrutiny not just on capability and safety, but on trust, transparency, and value-for-money, especially from a paying user base being asked to underwrite frontier AI development costs.
Read original article →