Ghost

Core is a screenless desktop computer built to run personal AI agents locally in your home, aimed at people who want their assistant, context, and data to live on hardware they own rather than in someone else's cloud. The device features an Nvidia RTX Pro 4000 SFF Blackwell GPU, proprietary software architecture, and a custom file system, all optimized to continuously process tasks without constant user prompting. It ships preloaded with three open-weight models, Qwen-3.8-Next, Qwen-3.8-27B, and Gemma-4-31B, while supporting additional models from Hugging Face or custom user training. Because there is no display, Core is accessed by a dedicated smartphone and web application, which serves as the control center from where users can message their personal agents or speak to them by voice , and the on-device models continuously gather context from apps, files, wearables, and smart-home devices. The launch matters now because Ghost is one of the first consumer attempts to package a serious local-inference rig as a finished product rather than a DIY build, with a clear pitch against general-purpose boxes. Javaid told TechCrunch that Core's hardware and software were built specifically for running AI agents, with models already installed, while a Mac Mini is a general-purpose computer that can be configured for AI. At $3,499, with no cloud and no subscription , and 30-day returns , it is positioned for early adopters who want always-on agents without recurring API costs or data leaving the house. Ghost is based in San Francisco and emerging from stealth alongside this launch. Javaid co-founded Ghost with Nicholas Chua (19), Yifei Chen (24), and Gautam Sharda (23) , and the team built the product from scratch, handling both the hardware design and the software stack in-house. The company raised an $11 million seed led by Andreessen Horowitz, with Abstract, Audacious Ventures, SV Angel and Nova joining.