Having agents operate everyday apps may be a false premise, argues viral Chinese thread
sujingshen · x · 2026-10-15
A Chinese tech thread argues that letting agents operate everyday apps like Taobao or WeChat may be a false proposition. Tasks like hailing a ride or ordering food take under 5 minutes — describing them to an AI assistant takes as long as doing them yourself; the litmus test: would you delegate this to a human assistant?
Many content and e-commerce apps are inherently kill-time products where users enjoy browsing and comparing. An experiment: an agent picked the top 10 products for a user, who rejected them all, then browsed Taobao manually and ultimately chose something from that same top 10 — users' expectations (good-looking, cheap, no bad reviews) only recalibrate through browsing itself. Widespread agent takeover would also gut app ad revenue.
The authors contend assistants should first fix their own ecosystems: Xiaomi's speaker fails on near-miss voice commands, and Xiaomi TV's Bilibili casting ignores voice "pause" while the remote works. The endgame may not be agents on existing apps but an entirely new, agent-native two-sided network, much as mobile replaced the PC web.
More from AGI Musings
- Every's Dan Shipper declares "the newsletter is dead" as persistent agents take over the inbox — danshipper · 2026-10-15
- Stockholm University compares AI-allowed data projects vs paper exam scores, 2022-2026 — paulnovosad · 2026-10-15
- Swartzism v0.0.2: packaging Aaron Swartz's access-to-knowledge doctrine for the post-AI era — kingslyj · 2026-10-15
- New essay argues consciousness is a journey living systems do — not AI — AnnaCiaunica · 2026-10-15
- Oversubscribed London AI x Science hackathon maps researchers' biggest software bottlenecks — ersatzben · 2026-10-15
- Hannah Fry: How OpenAI has triggered an existential crisis in mathematics — AnnaCiaunica · 2026-10-15