Work · Mo Zhang

Product 01

Token Plan

The world’s first token compute service from a provincial carrier

The intelligence of large models is fast becoming a new kind of utility — like water, electricity, and gas. But unlike those, tokens are still unfamiliar to most people: even after openClaw went viral in early 2026, few can say what the “intelligence” they consume really is.

Serving and educating the public is what a carrier is for, and its vast retail network is a ready-made classroom. We set out to turn compute into a ready-to-use token service: at Guangdong Mobile (the largest provincial carrier in China, and perhaps the world), we launched Token Plan first and have iterated ever since — it now offers stable access to the world’s leading models.

None of it came easy: the models and compute at the bottom, the aggregation and distribution platform in the middle, the plans and entry points on top — the team built every layer piece by piece. Tokens still have a long way to go before they reach everyone, but this is undeniably a good start. Come give it a try!

Poster of Guangdong Mobile’s Token Plan compute packages — Token trial plans launched in April 2026.

Smart Meeting Assistant

A team’s on-premises intelligence hub — data never leaves your domain

The most tiring part of meetings isn’t just the back-and-forth — it’s writing up the minutes afterwards. This little box in the meeting room takes care of that, and it’s foolproof: press the button to start, press again when the meeting ends, and minutes later a tidy summary arrives — who said what, what was decided, who does what next — ready to broadcast in one click.

It lives entirely on-site: transcription, speaker identification, and minutes generation all run on edge compute, robust enough to handle a 10-person meeting running six hours straight. Nothing confidential ever leaves the room — which is more than most cloud-based meeting assistants can say. Even users with the strictest data-privacy requirements can use it with peace of mind.

Wearable Gait-Monitoring Device

24/7 patient monitoring that misses nothing

Did you know that, clinically, one in three older adults has a gait disorder? For their day-to-day care, continuous gait monitoring is essential — but you can’t ask them to trek to the hospital every day.

We built a wearable from the chips, sensors, and battery up. Worn unnoticed on the lower back, it reads gait in real time with AI: steady or not, shuffling or not — and flags anomalies the moment they appear, no waiting for the next clinic visit.

The hard part was making it small: the algorithm had to fit into a low-power sensor chip and run around the clock. We polished our Edge AI until high-accuracy monitoring ran on a tiny power budget. Still, here’s hoping you never need it.