After spending time using mathematical models and data to steer financial investments, nineteen-year-old Zain Javaid found himself looking for a deeper challenge. The public debut of ChatGPT in 2022 fundamentally shifted his ambitions, steering his attention toward the untapped potential of personal artificial intelligence. That enthusiasm culminated in the creation of Ghost, a startup that has formally emerged from stealth mode following twelve months of development. Backing the initiative, venture capital firm Andreessen Horowitz led an $11 million seed funding round, joined by participants Abstract, Audacious Ventures, SV Angel, and Nova. Javaid is accompanied on the founding roster by co-founders Nicholas Chua, 19, Yifei Chen, 24, and Gautam Sharda, 23.
Engineered Hardware for Autonomous Digital Agents
Ghost has debuted its flagship product under the name Core, a specialized personal desktop machine crafted from the ground up to support autonomous software agents acting on an individual's behalf. While running complex neural networks on conventional home setups often results in operational friction and poor efficiency, Core addresses the issue by pulling a user's entire computing ecosystem together. It consolidates information spanning active desktop environments, mobile software, and connected smart home devices into a single piece of hardware designed to operate persistently without requiring continuous manual prompts.
Integrated Silicon and Component Specifications
Inside the chassis, Core packs an Nvidia RTX Pro 4000 SFF Blackwell GPU, a dedicated graphics processor designed to handle the heavy computational load associated with local model execution. Rather than requiring buyers to source high-end graphics chips separately, the processing hardware comes fully integrated into the package. Ghost has priced the system at $3,499. Advance preorders for the initial production run commenced on Monday, with customer fulfillment scheduled to begin during the final week of October. To support non-stop workflow management, the machine incorporates a custom software stack, a dedicated browser, and a tailored filing system.
Screen-Free Architecture and Ready-to-Run Models
Described by its creator as a self-contained brain in a box, Core dispenses with an integrated display panel and relies instead on mobile devices for user interaction. Individuals interface with the system through a companion smartphone application. While the platform supports voice interactions, users route their spoken words through the mobile app rather than addressing the physical box directly. Out of the box, Core arrives with three preloaded open-weight systems: Qwen-3.8-Next, Qwen-3.8-27B, and Gemma-4-31B. Owners retain complete flexibility to import alternative builds from the Hugging Face model repository or load custom checkpoints they have trained independently.
Distinction from Standard Compact Desktops
Addressing comparisons to compact desktop computers like Apple's Mac Mini, which enthusiasts often repurpose for local machine learning experiments, the company emphasizes the architectural divide. Where the Mac Mini operates as general-purpose consumer hardware adapted through third-party configurations, Core integrates its operating software and chip architecture solely around agent workloads. Ghost manages future updates through an over-the-air mechanism reminiscent of modern electric vehicles. Whenever upgraded model weights appear, the platform prompts owners regarding potential transitions, while an optional auto-update toggle enables hands-free adoption of subsequent model generations.
On-Premise Privacy and Cloud Independence
Manufacturing of the systems takes place directly at Ghost's operational facility in San Francisco. The strategic focus centers on delivering an appliance that eliminates cumbersome installation workflows. Because personal memories and neural weights reside strictly on the local drive, user data undergoes thorough encryption to prevent external entities, cloud hosting companies, and Ghost's internal team from gaining unauthorized visibility. The system keeps all underlying logic, codebase components, and mathematical weights on the device itself, ensuring full functionality even if Ghost ceases business operations in the future.
Outbound Network Safeguards and Contextual Intelligence
To supervise how autonomous agents utilize sensitive user credentials such as accounts and passwords, the engineering team is implementing a custom internal firewall. This defensive layer scrutinizes every departing network transmission leaving the system. If an autonomous process attempts an unapproved action or tries exfiltrating protected data outside the local perimeter, the firewall blocks the event and alerts the human owner for review. Demonstrating its contextual awareness, Core can proactively alert an individual about an impending illness by analyzing elevated resting heart rates, correlating late-night computer activity logged in browser records, and referencing prevailing flu season patterns in San Francisco.
