Why “Lobster” (龙虾) First Appeared on Computers — Not Phones

The curious lead

It has been reported that the new wave of AI agents — a phenomenon some outlets call “Lobster” (龙虾) — has taken root first on personal computers rather than on smartphones. Huxiu (虎嗅) republished a piece originally from ifanr (爱范儿) arguing that this is not an accident. Smartphones dominated the last decade as attention-grabbing, always-personal screens built for direct human interaction. But AI agents want to be always-on, sensing the world and acting without a human tapping the screen. So where do they land first? On machines that still speak “machine” as well as “human.”

Legacy architecture vs. locked-down convenience

The explanation is structural. Computers retain decades of “machine-first” design: command lines, file systems, installable tooling and user-configurable runtimes that let developers script, orchestrate and chain tasks. Phones, by contrast, were designed to be immediately intuitive — a jump straight to graphical, app-centric interfaces with strict sandboxing and tightly controlled permissions. That makes phones easier to use, but harder to repurpose as autonomous agents. It has been reported that Apple’s Mac mini became a hot item among developers seeking local compute for agents like OpenClaw; computers offer more compute headroom, persistent context and the ability to run scripts that talk to other programs in ways mobile apps generally cannot. And when “screenless” automation bumps up against app-store rules, simulated interaction apps have reportedly been blocked, pushing experimentation back to PCs.

Implications for AI on-device and geopolitics

Does this mean phones lose out forever? Not necessarily. But for now, the path to practical, autonomous assistants runs through the production-tool heritage of PCs. That has implications beyond user experience. Device form factors, software openness and even geopolitical supply-chain issues — such as global semiconductor constraints and export controls that shape where and how advanced inference runs — will affect whether sophisticated AI stays on clouds, moves to desktops, or finally arrives on edge devices in phones. For Western readers: think less about which gadget is cooler and more about which one preserves the hooks AI needs to do real work. The “Lobster” moment is a reminder that architecture and platform policy matter as much as models themselves.