Ghost Builds Core for the Personal AI Hardware Era

Ghost is building Core, a $3,499 personal AI computer designed to run local models, remember context over time, and act through the apps and devices around its owner. The San Francisco startup is led by founder and CEO Zain Javaid alongside founding-team members Nicholas Chua, Yifei Chen, and Gautam Sharda.

Core enters the consumer AI market with a clean but difficult proposition: the most useful personal agent will need intimate context, yet the more it knows, the harder it becomes to trust a remote service with the keys. Ghost is betting that dedicated hardware in the home can make ownership, memory, and control part of the product rather than a settings-page promise.

That bet now has $11M in Seed funding led by Andreessen Horowitz, with Abstract, Audacious Ventures, SV Angel, and Nova participating. The financing matters because Ghost is not shipping a lightweight app. It is taking on hardware, model management, integrations, security, and the daily behavior of an agent that is supposed to be useful without becoming reckless.

About Ghost and Core

Ghost emerged from stealth in October 2026 after roughly a year of building. Its first product, Core, is a screenless computer accessed through a phone or companion app. Ghost says Core connects to files, email, calendars, wearables, cameras, screen history, and smart-home devices, then uses that context to answer questions and perform tasks continuously.

The current machine includes an NVIDIA RTX PRO 4000 Blackwell SFF Edition GPU with 24 GB of memory, an AMD Ryzen 5 7600 processor, 64 GB of DDR5 memory, and 1 TB of NVMe storage. Core comes with local models installed and allows owners to change models without rebuilding their stored context. The price is $3,499, with no subscription advertised.

Those specifications are not the whole pitch. Plenty of technically capable people can run local models on a desktop or configure a small computer as an agent host. Ghost is selling the integration: hardware, models, memory, connections, permissions, updates, and a user experience meant to work without turning every owner into the unpaid systems administrator of their own robot butler.

Why Ghost Matters Right Now

Personal AI is moving from chat toward persistence. A chatbot waits for a prompt. An agent with memory can notice patterns, connect events across services, and act before someone writes a new instruction. That transition creates more utility, but it also creates a larger trust surface.

Ghost's answer is to keep conversational AI and memory on hardware the customer owns. The company's privacy policy says Core can process conversations, files, connected-app content, screen history, health information, financial information, and credentials on-device. It also says Ghost does not use that information to train AI models.

Local compute is meaningful, but it is not a magic cloak. Online actions still require networks and outside services. Copies of data and derived memories can persist on Core or paired devices. TechCrunch reported that Ghost was still developing a credential firewall to monitor outgoing requests at launch. For this category, trust will be won in implementation details, not in how often the word privacy appears on a product page.

The Team Building a Personal AI Appliance

Javaid is 19 and serves as Ghost's founder and CEO. Nicholas Chua is a founding-team member who says he leads growth. Yifei Chen joined as founding designer to shape the product, brand, experience, and design engineering. Gautam Sharda is also part of the founding team, although Ghost has not published a narrower role title that could be independently confirmed.

That small team is attempting an unusually broad build. Consumer hardware demands supply planning, assembly, support, returns, and reliable updates. Agent software demands model evaluation, memory design, integrations, permission systems, and protection against unwanted actions. Product design has to make all of that understandable to people who do not want to debug an AI appliance before breakfast.

Ghost's public setup guide offers an early view of that work. It moves owners through pairing, Wi-Fi, model selection, app and data connections, memory construction, and first use. The sequence is deliberately ordinary. That is a strategic requirement, not a cosmetic one. The winners in personal AI hardware will have to hide substantial technical complexity without hiding the controls users need.

The Market Test for Core

Core sits between several established categories. Cloud assistants offer access to powerful models without a $3,499 hardware purchase. General-purpose computers can run local models and keep improving. Open-source tools let advanced users assemble custom agent systems. Phones, wearables, and smart-home platforms already occupy many of the connection points Ghost wants to coordinate.

Ghost therefore has to prove that a dedicated appliance is more than an expensive container for components that can be bought elsewhere. The company needs a coherent experience that makes persistent context dependable, makes permissions legible, and makes model updates routine. It also needs to support the physical product after the opening-week curiosity wears off.

No public revenue, unit-sales, customer-count, or retention figure was available at publication. No independent security audit or comparative performance benchmark was found either. The first batch and late-October shipping plan create a near-term test, but real traction will be visible only after Core has spent time inside customers' homes.

What Ghost Signals for Consumer AI

Ghost reflects a broader industry shift from renting isolated moments of intelligence to owning systems that accumulate context. That shift changes the competitive questions. Model quality still matters, but memory, permissions, data location, physical reliability, and graceful failure matter too.

The company's strongest idea is that the architecture of trust can be a product feature. Its hardest problem is proving that local ownership produces an experience worth the premium while controlling what an always-on agent can see and do. If Core succeeds, personal AI may develop an appliance layer alongside cloud subscriptions and general-purpose computers.

Ghost has assembled funding, hardware, and a young cross-functional team around that possibility. There is no verified careers page or public open-role inventory yet, so the present signal is not a hiring campaign. It is the scope of the build itself: a small company is trying to make persistent AI feel like something people can own, understand, and keep in the home.

The category will not be settled by a launch video or a specification sheet. It will be settled by whether Core ships reliably, earns access carefully, improves with context, and stays useful after the novelty fades. Ghost has picked a product where intimacy is both the advantage and the liability. That is exactly why the company is worth watching.