AI手搓的Cowork“李鬼”版跟“李逵”一样能打，还免费？

Core Insights - Anthropic's Cowork is a desktop AI agent that allows users to automate tasks without programming, but it is expensive, available only to Max users at a minimum of $100 per month [1] - The rapid development of a free open-source version, OpenWork, within 48 hours indicates low technical barriers and clear product logic [1] - The development cycle of Cowork was only 10 days, with most of the code generated by AI, showcasing the potential for AI to create AI [1][9] Product Comparisons - Manus, developed by a company acquired by Meta, is known as the "first general AI agent" and achieved $100 million in annual recurring revenue within 8 months of its launch [3] - Gemini CLI, Google's open-source terminal agent, offers free access to Gemini 2.5 Pro and supports various integrations, but has a higher usage barrier due to its command-line interface [5][6] - ChatGPT Agent, launched in July 2025, operates in a virtual machine environment and has a lower baseline success rate of 12.5% in practical tests, indicating a need for optimization [5][6] Technical Architecture - Manus employs a multi-agent system using a MapReduce architecture, allowing it to handle large-scale tasks efficiently [7] - Cowork operates within a local folder using sandbox mechanisms for security, while Gemini CLI provides direct access to system terminals, offering flexibility but with higher risks [6][8] - The integration of multiple agents and tools represents different balances of security and capability across these products [7] Industry Implications - The emergence of AI building AI signifies a shift in software development timelines, reducing them from months to days [9] - The recursive improvement process within Anthropic has led to a significant increase in coding efficiency, with AI now handling 60% of coding tasks [10] - The transition from traditional software development roles to AI-assisted roles is reshaping the engineering landscape, with engineers focusing more on code review and architecture [12] Future Trends - The trend of AI constructing its successors is irreversible, with predictions indicating that by 2028, 90% of B2B procurement will be handled by AI agents [22] - The potential for AI to transform workflows into AI-first designs is significant, although challenges related to security and reliability remain [22][23] - The shift from passive chatbots to proactive AI agents represents a fundamental change in human-computer collaboration, with profound implications for productivity and task execution [23]