There is no shortage of startup accelerators, but there are not many projects that attract the close attention of investors, and PearX is one of them. PearX is a 12-week program that accepts a maximum of 20 startups per session, and it is operated by Pear VC, a venture capital firm focused on pre-seed and seed round investments.

PearX Held once every six months, demo, day, it almost always attracts top venture capitalists to attend, which also reflects that the overall quality of the selected companies has always been very high. Some of the companies that have emerged from recent sessions include Known, which uses voice AI to match people up for dates and received financing from Forerunner Ventures; there is also Andera, a startup that automates corporate auditing and compliance tasks, and this summer it received $37 million in Series A financing from Lightspeed.

There are quite a few differences between PearX and the pioneer of startup accelerators, Y Combinator. In addition to being much smaller in scale, it does not offer standard terms and conditions. The maximum investment amount can reach up to 2 million US dollars. Furthermore, unlike YC, some of the most highly regarded startups in YC receive financing before the end of their projects, while PearX claims to keep the identities of the participating companies confidential until demo day.

TechCrunch attended the latest demo of Pear in San Francisco last week, and then stayed behind to ask several venture capitalists which of these 16 startups they were most interested in.

The following are 5 of the companies that have sparked the most discussion.

Speridlabs

Its business: providing spatial base models for robots, games, and special effects.

Why it's getting attention: Speridlabs is not the only startup developing world models. Such models aim to provide similar capabilities for 3D space as large language models do for language. However, the company believes that more established competitors, including Runway, Odyssey, Google, and Genie, cannot be queried or modified. To address this issue, Speridlabs developed Mundus, which they call the "3D version of Midjourney," because when a part of it is modified, it can maintain the continuity of the geometric structure.

Saia

Its business: to develop a chip that is fast, cost-efficient, and capable of direct inference on device side.

Why it's getting attention: The GPU of Nvidia and the TPU of Google require expensive and power-consuming memory, which is currently in short supply. Saia claims to have designed a chip that bypasses traditional memory and runs directly from flash storage. The company says this chip is significantly faster, with a capacity eight times that of the latter, and yet it consumes only a quarter of the power. Jetson is a leading local AI chip and development board. Saia states that the company is already in discussions with Samsung regarding memory integration, plans to start manufacturing test chips next year, and hopes to achieve mass production by 2028. Hardware development has always been exceptionally challenging, but the 20-year-old founder, Ayaan Govil, managed to convince Pear VC, the co-founder Mar Hershenson – a semiconductor engineer with a doctorate in circuit design – to bet on his vision.

Ren

Its business: a safer AI personal assistant.

Why it's getting attention: Ren claims that its functionality is similar to that of Muse and Instinct, but it places more emphasis on security and privacy. The company states that it tries to keep data on the device side as much as possible, or uses secure private clouds, and performs operations according to strict boundaries set by users. In addition, Ren stands out from its competitors through a fully automated voice system; the latter usually rely on human operators to make calls. Ren indicates that this approach completely eliminates human involvement, thereby protecting user privacy.

Veros

Its business: AI native trust and estate planning tools.

Why it's getting attention: Setting up a trust is both expensive and time-consuming. Veros hopes to simplify the tasks typically performed by lawyers, wealth managers, and trust administrators through AI, which include recommending structures and managing assets throughout their life cycle. The company currently manages $250 million in AUM and is also applying for a trust license in order to operate as a regulated trust company in the future.

Datum

Its business: targeting AI engineers in industrial design.

Why it's getting attention: Datum aims to accelerate the development of physical products by helping engineers find and reuse previously created content. The company will index its enterprise's 3D design library and utilize its proprietary Geometric Fingerprint technology to identify parts based on their shape, thereby simplifying and automating engineering processes in the vast market of physical designs.