36Kr Analysis: OpenClaw Agents May Signal the End of the Traditional App Era
What Happened
Chinese tech publication 36Kr published a detailed analysis arguing that OpenClaw represents a "1995 moment" — comparable to Windows 95's revolutionary impact — where AI agents are poised to replace traditional mobile and web applications as the primary interface between users and digital services. The article proposes a fundamental architectural shift: from the current model of "User → Application Interface" to a new paradigm of "User → AI Agent → Application Service."
The analysis draws on data from OpenRouter and a16z showing that agent-driven output tokens have now exceeded half of total platform output, suggesting the transition from human-driven to agent-driven API consumption is already measurably underway. According to 36Kr, this shift means apps will gradually degenerate from user-facing products into "data and action interfaces for agents" — users won't interact with apps directly but instead have their AI agents call services on their behalf.
The piece tracks how every major Chinese tech company is positioning for this shift: Tencent released five AI agent products simultaneously, ByteDance launched ArkClaw via Volcano Engine, Baidu built the "Red Finger Operator" for cross-app automation, and Alibaba is building enterprise agents through DingTalk.
Why It Matters
The framing matters as much as the substance. When China's most influential tech publication declares the app era potentially over — and backs it with platform-level consumption data from a16z — it shifts the Overton window for how enterprises and developers think about product strategy. If the a16z/OpenRouter data is directionally correct, we are already past the tipping point where more AI-generated API calls are made by agents than by humans. This has profound implications for app monetization (ads become less relevant when agents bypass UI), API design (optimizing for agent consumption rather than human browsing), and platform economics (the agent layer, not the app layer, becomes the new distribution bottleneck).
What's Next
The key test will be whether consumer behavior actually follows the thesis. Currently, agent-driven token consumption is concentrated among developers and power users. The question is whether mainstream users — especially in China, where super-apps like WeChat dominate — will delegate daily tasks to agents within the next 12-18 months. Watch for app download trends and daily active user metrics from major Chinese platforms as leading indicators of whether the "post-app" era is materializing or still aspirational.