Maniformer launched Mifengpai in Shanghai on September 23, 2026, with a straightforward idea: use wearable capture devices and a public task platform to record how people do real work, then process those recordings into data for robotics. CEO Yao Maoqing has argued that robots need more real human-motion data; the platform is Maniformer’s proposed way to collect it at scale. 30

How Mifengpai works

The platform connects data-collection equipment and software with task providers, real-world settings and participants. In practice, users choose a task in the app, record the requested activity with MEgo equipment, submit the recording for review and receive payment based on approved usable data time. 192530

Tasks described in launch coverage include everyday actions such as folding clothes, making tea and arranging goods. The recordings are intended to capture human actions in the environments where those actions normally happen, rather than relying only on demonstrations performed in a robotics lab. 30

What the MEgo devices and data engine do

Mifengpai’s collection system uses MEgo devices alongside a mobile app and MEgo Engine. Launch reporting describes MEgo View as a head-worn device with camera coverage exceeding 300 degrees, while MEgo Gripper is reported to reconstruct manipulation movement at millimeter scale. 230

The aim is to capture more than a basic video. Product and launch coverage describes collecting visual information alongside movement and interaction data; MEgo Engine then processes, labels and evaluates recordings so they can be prepared for model training or evaluation. 253645

This distinction matters to the platform’s model: people produce demonstrations in real settings, while the equipment and processing software are meant to turn those demonstrations into structured data. The company describes Mifengpai as the link between those contributors, work sites and robotics data needs. 30

What early participation figures show

Launch coverage reported that the one-month beta had 20,000 registered users and 13,000 task submissions. Those figures indicate early participation, but they do not establish how many submissions passed review, how much usable data was produced or whether the resulting data improved robot performance. 1941

Other reporting lists 22 categories, more than 5,000 task types and over 50,000 real-world settings. These are reported platform figures, not independent measures of data quality or robot capability. 714

What remains unproven

Mifengpai offers a method for gathering human demonstrations and paying for recordings that pass review. But the launch evidence describes the platform’s design and reported early activity; it does not independently demonstrate that the data has made robots more capable, or verify outcomes such as long-term earnings and the quality of accepted submissions. The distinction is important: collecting data is a proposed response to a bottleneck, not proof that the bottleneck has been solved.

Maniformer launched Mifengpai in Shanghai on September 23, 2026, with a straightforward idea: use wearable capture devices and a public task platform to record how people do real work, then process those recordings into data for robotics. CEO Yao Maoqing has argued that robots need more real human-motion data; the platform is Maniformer’s proposed way to collect it at scale. 30

How Mifengpai works

The platform connects data-collection equipment and software with task providers, real-world settings and participants. In practice, users choose a task in the app, record the requested activity with MEgo equipment, submit the recording for review and receive payment based on approved usable data time. 192530

Tasks described in launch coverage include everyday actions such as folding clothes, making tea and arranging goods. The recordings are intended to capture human actions in the environments where those actions normally happen, rather than relying only on demonstrations performed in a robotics lab. 30

What the MEgo devices and data engine do

Mifengpai’s collection system uses MEgo devices alongside a mobile app and MEgo Engine. Launch reporting describes MEgo View as a head-worn device with camera coverage exceeding 300 degrees, while MEgo Gripper is reported to reconstruct manipulation movement at millimeter scale. 230

The aim is to capture more than a basic video. Product and launch coverage describes collecting visual information alongside movement and interaction data; MEgo Engine then processes, labels and evaluates recordings so they can be prepared for model training or evaluation. 253645

This distinction matters to the platform’s model: people produce demonstrations in real settings, while the equipment and processing software are meant to turn those demonstrations into structured data. The company describes Mifengpai as the link between those contributors, work sites and robotics data needs. 30

What early participation figures show

Launch coverage reported that the one-month beta had 20,000 registered users and 13,000 task submissions. Those figures indicate early participation, but they do not establish how many submissions passed review, how much usable data was produced or whether the resulting data improved robot performance. 1941

Other reporting lists 22 categories, more than 5,000 task types and over 50,000 real-world settings. These are reported platform figures, not independent measures of data quality or robot capability. 714

What remains unproven

Mifengpai offers a method for gathering human demonstrations and paying for recordings that pass review. But the launch evidence describes the platform’s design and reported early activity; it does not independently demonstrate that the data has made robots more capable, or verify outcomes such as long-term earnings and the quality of accepted submissions. The distinction is important: collecting data is a proposed response to a bottleneck, not proof that the bottleneck has been solved.