Hub

What it does

Hub.xyz provides custom real-world data created, annotated, and curated by a global network of retail users and professionals. The platform offers off-the-shelf datasets across four modalities—egocentric video, image, video, and audio—as well as bespoke collection services for long-tail data projects. Data is pipeline-ready, structured, labeled, and annotated, with features like IMU data, RGB depth, 3D hand reconstruction, speaker diarization, and emotion labels.

Who it is for

Hub serves frontier AI labs, robotics companies, and small & medium businesses (SMBs). AI labs and robotics firms can purchase ready-made or custom datasets to train physical AI models. SMBs can join Hub's verified network to turn downtime into income by capturing real-world data. The network includes creative studios, pharmacies, medical clinics, hair salons, cafés, bakeries, restaurants, wood workshops, warehouses, production facilities, and schools.

Why it matters

Training physical AI—robots, autonomous systems, and embodied agents—requires diverse, high-quality real-world data that is often scarce or expensive to collect. Hub leverages a human network across 150 countries to generate data that "don't exist yet," enabling faster iteration and more robust models. The platform's end-to-end pipeline (brief to delivery in ~36 hours) reduces the time and cost of data collection, making it accessible to both large labs and smaller businesses.

Launch signal

Hub is backed by Y Combinator (P26) and has an active community of 240K followers on X and 235K Discord members. The platform reports over 18.7 million total contributions from its network. Off-the-shelf datasets include 54k hours of egocentric video, 2.4M images, 70k video clips (18k hours), and 80k hours of audio in 47 languages. The bespoke collection service promises quotes within hours and delivery within 36 hours.

Brand and naming

The name "Hub" is short, memorable, and suggests a central connection point between data contributors and AI builders. The .xyz domain reinforces a modern, tech-forward positioning. The tagline "Real-world data to train physical AI" clearly communicates the value proposition, while the website's emphasis on "human to machine symbiosis" and "built for the machines that build everything else" positions Hub as an essential infrastructure layer for the AI economy.