The voice mega-app

In China, WeChat is not an app: it is infrastructure. Messaging, payments, taxis, doctor's appointments, bills: everything goes through its mini-programs, without ever leaving the interface. The West never had an equivalent. The American giants carved up the pie, and every service kept its own app, its own account and its own notifications.

I believe this Western mega-app will eventually exist, but in an unexpected form: a voice agent. It won't own any service, it will speak to each of them on our behalf.

Voice first, screen as a last resort

For all communication and productive actions, voice becomes the main interface. The phone stays in the pocket, screen off, and the agent runs in the background at all times. You only unlock the screen for two things: entertainment, and critical actions, such as a transfer in your banking app.

For this to work, the agent has to speak first: "Your train leaves from platform 7 in twenty minutes." "The free trial ends tomorrow, shall I cancel?" And above all, it has to know when to stay quiet: filter, group, only interrupt when it matters. Along the way, the whole attention economy model starts to wobble. When nobody looks at their screen anymore, notifications have no surface left to capture our attention.

The app wall

The first full-scale attempt took place in December 2025, with ByteDance's Doubao phone. Its assistant simulated taps on the screen to operate apps on the user's behalf. Within days, WeChat, Alipay, Taobao and several banks restricted or blocked it. The reasoning is simple: a payment system that can no longer tell whether a human or an AI is in front of the screen sees its entire security collapse.

The lesson is clear: operating other people's interfaces is a bridge, not a foundation. The future belongs to protocols, where the agent sends structured instructions directly to services. WeChat has in fact started opening up in that direction with several phone makers.

That leaves trust, which cannot be decreed. Read, draft, send, pay: each level of autonomy is earned one action at a time, with a readable history and the ability to undo. It is this gradient that will shrink, year after year, the list of things we still take our phone out for.