19-Year-Old's Ghost Raises $11M for Local AI 'Brain in a Box'

Overview

Zain Javaid, a 19-year-old founder, has emerged from stealth with Ghost, a startup dedicated to creating hardware for personal AI agents. The company announced an $11 million seed round led by Andreessen Horowitz on Monday, alongside participation from Abstract, Audacious Ventures, SV Angel, and Nova.

Ghost’s flagship product, named Core, is a $3,499 desktop computer designed specifically to run AI models locally. Unlike general-purpose machines, Core aims to centralize a user’s digital life—spanning desktop files, applications, and smart home devices—into a single piece of hardware that continuously operates AI assistants.

“You can run models on existing hardware, but it’s a terrible experience on many levels.”

Javaid, who serves as the CEO of Core, emphasized that the device is engineered to remove the friction typically associated with local AI deployment. Preorders for the first batch open this week, with shipping scheduled for the last week of October.

The Origin of Ghost

Before founding Ghost, Javaid worked as a quant, using mathematics and data to guide investment decisions. However, he found the work unfulfilling. That changed in 2022 with the release of ChatGPT, which sparked a fundamental excitement in him that he had never experienced before.

His interest in personal AI quickly evolved into action. After a year of building in stealth, Ghost now includes a founding team composed of peers and young engineers: Nicholas Chua, 19; Yifei Chen, 24; and Gautam Sharda, 23. The collective experience of the team centers on bridging the gap between powerful AI models and practical, localized hardware solutions.

Javaid noted that while his background in quantitative analysis provided a strong foundation in data handling, it was the emergence of large language models that redirected his professional ambitions toward building infrastructure for autonomous agents.

Core: The 'Brain in a Box'

Core is marketed as a “brain in a box” without a screen. Users interact with the device through a smartphone or dedicated app, with voice support available via the app rather than hardware microphones on the unit itself. The machine comes preinstalled with three prominent models: Qwen-3.8-Next, Qwen-3.8-27B, and Gemma-4-31B.

The hardware specifications are tailored for AI workloads. Core includes an Nvidia RTX Pro 4000 SFF Blackwell GPU, which is bundled into the base price. This eliminates the need for customers to purchase expensive discrete graphics cards separately, a common barrier to entry for running local AI.

Local Processing and Privacy

A primary selling point of Core is its commitment to privacy and local processing. All personal information, model weights, and memory are stored and encrypted on the device. Ghost states that no cloud provider or third party, including the company itself, can meaningfully access this data.

“All of the source code, logic, and model weights live on the device, not the services.”

Javaid explained that the device is programmed to remain functional even if Ghost ceases to exist as a company. This design philosophy ensures that users retain full ownership of their AI ecosystem, independent of any external service maintenance.

The system also features a custom firewall that monitors every outgoing network request. If an agent attempts an unsanctioned action or tries to exfiltrate data, the firewall rejects the request and flags it for user review. This adds a layer of security to prevent autonomous agents from misusing user credentials or exposing sensitive information.

Software and Updates

Ghost differentiates Core from general-purpose computers like the Apple Mac Mini by emphasizing its specialized software layer. While a Mac Mini can be configured for AI, Core is built from the ground up for it, featuring its own operating system, browser, and file management system.

The company compares its update model to Tesla’s approach. New models or improvements can be pushed over-the-air. Users are given the choice to switch to new model generations manually or to enable auto-updates, ensuring they are always running the latest capabilities.

Additionally, users have granular control over how much autonomy they grant their Core agents. The system is designed to be proactive, potentially sending notifications based on aggregated context. For example, Core might alert a user that they could be getting sick by correlating an elevated resting heart rate, late-night computer activity, and seasonal flu trends in their location.

Market Position and Competition

Despite the involvement of tech giants in the AI space, Javaid expressed confidence in Ghost’s ability to carve out a niche. He argued that incumbents often receive credit merely for assembling existing ingredients, whereas Ghost is tackling the complex software challenges of reasoning, memory, and proactive action.

“If we can’t do that, Apple isn’t the reason we failed.”

Javaid credited Apple with creating excellent hardware and driving disruption in consumer technology. However, he asserted that the core challenge for Core lies in the software layer that allows AI to understand and navigate a user’s life effectively.

The company is currently manufacturing units at its San Francisco facility. The goal is to deliver a fully functioning computer that requires no complex setup or configuration, appealing to users who want the power of AI without the technical hurdles.

With the seed round complete, Ghost is focused on delivering the first batch of Cores to early adopters. The startup aims to prove that dedicated local AI hardware can offer a superior experience compared to repurposed general-purpose PCs, setting the stage for future growth in the personal AI agent market.