PearX’s latest demo day highlighted five startups that investors said were among the most interesting of the 16 companies in the latest cohort. Operated by Pear VC, PearX is a 12-week program that caps each cohort at 20 companies, with an investment focus on the pre-seed and seed stages.
The event was held in San Francisco last week and attended by prominent investors, while PearX keeps participating companies confidential until demo day. The program differs from Y Combinator through its smaller size and nonstandardized terms, as its investments can reach $2 million.
Companies in AI and hardware
Speridlabs is developing spatial foundation models for robotics, gaming, and special-effects applications. The company says its Mundus platform resembles a “3D Midjourney,” while keeping the scene’s geometry fixed when one of its elements is modified. It believes that some competitors, such as Runway, Odyssey, and Google Genie, do not allow models to be queried or modified in the same way.
Saia is working on a chip to run AI inference directly on the device. It claims that its design eliminates the need for traditional memory by running models from flash storage, with higher speed, eight times the capacity, and four times lower power consumption compared with an Nvidia Jetson board. The company is discussing integrating memory with Samsung and plans to begin manufacturing test chips next year, targeting commercial production in 2028.
Privacy and financial services
Ren is developing a personal AI assistant focused on security and privacy. It says it keeps data on the device whenever possible or uses a private cloud, and reviews actions according to controls defined by the user. The company also relies on a fully automated voice system to make calls instead of using human operators.
Veros offers what it describes as an AI-powered credit and estate-management planner. It aims to simplify the work of lawyers, wealth managers, and estate administrators, including proposing structures and managing assets over the long term. According to the article, the company manages $250 million in assets and is seeking a trust charter that would allow it to operate as a regulated trust company.
Industrial design automation
Datum is developing an AI engineering tool designed for industrial design. The platform indexes the company’s library of 3D designs and then uses its proprietary Geometric Fingerprint technology to identify parts based on their shape, with the goal of helping engineers find and reuse previous work.
Why does this demo day matter?
The five companies reflect different paths for applying AI beyond general-purpose conversational models: spatial world modeling, on-device inference, privacy protection, trust services, and industrial engineering. However, most of the figures and capabilities mentioned are based on the companies’ own claims, and the article does not provide independent test results. Saia’s manufacturing plans and Veros’s application for a trust charter also represent future stages that have not yet been achieved.