Observed Signal · Jul 27, 2026 · Partnership · Source: techcrunch · Impact: 2/5 · Sentiment: Positive
Encord trials brain-wave data to train physical AI
Encord is running trials in a San Leandro warehouse to create new training data for robotics by combining egocentric video, arm muscle sensors, leader-follower rigs and brain-wave measurements. The brain-wave headset used in the trial was built by German startup Zander Labs; Encord plans to tag datasets with mental-state signals (error, intent, surprise) to test whether they improve robot model performance. Encord says physical training data is scarce and expensive to produce, driving a business model around manufacturing datasets for manipulation tasks. Company leaders estimate the scale needed to advance physical AI is orders of magnitude larger than current corpora, and Encord is experimenting with dense annotation and multimodal signals to increase dataset value.
Highlights a specific industry response to the physical-AI data bottleneck—new multimodal dataset generation and dense annotation—which matters to robotics and AI training businesses but is not a major platform-level or regulatory shift.
Track OpenAI Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Encord is running a trial that records pilots' brain waves while they perform manipulation tasks to create brain-wave-tagged training data.
- The brain-wave headset used in Encord's trial was built by Zander Labs, a German neuroscience startup.
- Encord collects egocentric video, remote robot-operated data and uses leader-follower rigs to manufacture physical training datasets.
- Vineeth Velmurugan, Encord’s head of robot learning, said breaking through in physical AI will require a dataset roughly five times the size of YouTube’s video corpus.
- Several pilots at Encord previously worked at Scale, an AI data annotation firm.
Connected Companies & Entities
2 Entities mapped“A veteran of OpenAI’s robot lab and Berkshire Grey, the warehouse automation firm, Velmurugan joined Encord to build the company’s internal ...”
“When TechCrunch visited, pilots were using leader-follower rigs — paired robotic arms, one controlled directly by a human operator and one t...”
Ontology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Human Archive Taps India's Gig Workers to Train Robots
Human Archive, a Silicon Valley startup founded by four students from Berkeley and Stanford (Samay Mani, Rushil Agarwal, Shloke Patel, Raj Patel), is collecting large-scale egocentric and multi-sensor data from India’s gig-economy workers to train physical AI and robotics systems. The company says it has deployed more than 1,000 headsets and 50+ devices and is developing custom hardware (including tactile gloves, wrist cameras, and full-body motion capture) to capture synchronized RGB-D, motion, and force data. Human Archive raised $8.2 million from investors including Wing Venture Capital, NVP Capital and Y Combinator, plus angels from OpenAI, Nvidia, Google, Meta and others. The effort has drawn privacy scrutiny and some rejections from major home‑services firms (Pronto, Urban Company) even as the startup partners with smaller providers and pilots expansion to Southeast Asia and the U.S.
Cerebionics Develops Non-Invasive BCI to Translate Intent into Machine Commands
Cerebionics, a Norwegian startup founded by Agnessa Pedersen, is developing a non-invasive brain-computer interface (BCI) platform that uses EEG electrodes to translate human intent into commands for machines. The platform aims to provide an additional control channel for various systems, including drones, unmanned ground vehicles, and C2 APIs. Unlike Neuralink's invasive approach, Cerebionics focuses on higher-level commands rather than fine motor control. The startup has tested its technology with Ukrainian soldiers, receiving feedback to improve robustness. Cerebionics has received commercial interest through letters of intent and is currently scaling its team. The company anticipates applications in defense, healthcare, consumer, and industry, with future potential in gaming and integration into wearables like smart glasses.
Physical AI: Humanoid Robots Are Taking Off
Advances in hardware (motors, microelectronics) and software (generative AI, world models) are accelerating the commercial adoption of humanoid robots — often called "Physical AI" — across retail, hospitality, healthcare and logistics. The article highlights a symbolic win by Honor’s remotely controlled humanoid runner "Blitz" at the Beijing half‑marathon and cites a Roland Berger study projecting $300–750 billion in humanoid‑robot revenue by 2035 (with a long‑term Physical AI market potential of $4 trillion). Germany ranks third among system integrators after China and the USA, with local players such as Agile Robots, Neura Robotics and integrator RoboPlanet deploying test cases. Practical barriers remain: autonomy, training data (Neura’s "NeuraGyms"), facility integration (e.g., elevators, doors), insurance and onsite readiness. Vendors are also exploring branding use cases where robotic movement and choreography become part of a brand’s identity.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.