Ghost's $3,499 Screenless Box Wants to Run Your AI Agents 24/7 — and a16z Just Bet $11M On It

A 19-year-old, $11M, and a box with no screen

Zain Javaid is 19 years old. He just raised an $11 million seed round led by Andreessen Horowitz. And the product he's selling is a $3,499 computer with no monitor, no keyboard, and no conventional interface — because it wasn't built for humans to stare at.

It was built for AI agents to run on.

Javaid's company, Ghost, emerged from stealth this week with its flagship product: Core, a dedicated personal computer whose sole purpose is to host and run personal AI agents continuously, without interruption, without a screen, and without requiring you to babysit it.

The pitch is deceptively simple: your AI agents shouldn't have to share compute with your browser tabs and Slack notifications. They deserve their own hardware.

What Core actually is

Core is a purpose-built machine designed to run AI agents as persistent, always-on processes. Think of it less like a laptop and more like a server you keep at home or in an office — except it's designed specifically around the workflows of agentic AI rather than traditional computing tasks.

The device runs agents that can handle scheduling, research, trading, communication, and other tasks that benefit from continuous operation rather than being triggered only when you open an app. The premise is that most AI tools today are reactive — you prompt, they respond. Core is built for AI that acts proactively, on a loop, while you're doing something else entirely.

At $3,499, it's not cheap. But the target customer isn't someone who wants a smarter Alexa. It's founders, operators, and builders who are already running complex AI workflows and losing compute and context every time their laptop sleeps.

From quant trading to personal AI infrastructure

Javaid reportedly spent years wanting to work in quantitative trading before pivoting hard into AI infrastructure. That background matters — quant trading is one of the few fields where always-on, autonomous, rule-following software agents have been standard for decades. Algo trading desks don't pause because someone closed a laptop lid.

Javaid is essentially asking: why should your personal AI?

The logic maps cleanly onto where AI-native companies are heading in 2026. The pattern across the fastest-moving solo founders and small teams isn't better prompting — it's persistent, automated systems that run without human input for hours or days at a time. Core is hardware infrastructure for exactly that operating model.

Why a16z wrote the check

Andreessen Horowitz has been direct about its thesis on agentic AI: the next wave isn't chatbots, it's agents that complete multi-step tasks autonomously. The firm has backed a string of companies building in that direction, from agent orchestration platforms to vertical-specific automation tools.

Ghost fits that thesis at the infrastructure layer. Rather than building another agent framework or LLM wrapper, Javaid is selling the physical substrate — the always-on compute that makes persistent agents practical for individuals rather than just enterprises with server budgets.

The $11M seed also signals that a16z sees hardware as a credible play in the agentic stack, not just software. That's a meaningful conviction to put capital behind at seed stage.

The solo founder angle

Core's most obvious early adopters aren't enterprises. They're the kind of people this site tracks closely: solo founders running AI-native operations with minimal headcount and maximum automation.

If your company runs on agents — for marketing, research, customer operations, lead generation — the bottleneck isn't usually the model. It's continuity. Agents that restart every session, lose context, or compete for resources with your other open applications are slower and less reliable than they need to be.

A dedicated machine changes the economics. It's a fixed infrastructure cost ($3,499, plus whatever the subscription model looks like) rather than a recurring compute cost that scales with usage. For a founder running five or six concurrent agent workflows, that math can work.

This connects directly to the revenue per employee metric that defines real AI-native performance. Every persistent agent running on Core is, in effect, a non-human worker clocking hours while the founder does something else. The ones who build this kind of infrastructure soonest will widen the gap fastest.

What it means for the agent stack

Ghost isn't alone in recognizing that agentic AI needs better infrastructure. The software side has been crowded for two years — orchestration layers, memory systems, multi-agent frameworks. Hardware has been quieter.

But the logic is sound: if agents are going to be as important as Ghost and a16z believe, the current model of running them as processes on a general-purpose laptop is going to look as crude as running a production database on your home desktop.

Dedicated agent hardware is a natural step. The question is whether $3,499 is the right price point to drive early adoption, or whether Ghost needs enterprise traction first to subsidize a consumer price drop later.

For context, the Mac Mini starts at $599. A Mac Studio sits around $2,000. Core costs more than both — without a display — so the value proposition has to be about the software layer, the agent runtime, and the always-on reliability, not the raw compute specs.

What to watch

Ghost is pre-revenue in any meaningful public sense, which means the $11M is buying time to prove out the hardware-software bundle before the next raise. A few things will determine whether Core becomes infrastructure or a novelty:

• Agent reliability: Does Core's dedicated environment meaningfully outperform a well-configured cloud VM for persistent agents? If not, the hardware moat evaporates.
• Developer ecosystem: Agent hardware without an active builder community is just an expensive box. Ghost needs developers building specifically for Core's runtime.
• Pricing model: The $3,499 figure is for the device. What does the ongoing subscription look like? That's where the real business is.
• Founder credibility at scale: Javaid is 19, which the press keeps noting. It's largely irrelevant to the product — but it means Ghost's execution track record is still being written.

For founders building lean AI-native companies, Core is worth watching as both a product and a signal. The fact that a16z backed dedicated agent hardware at seed tells you something about where the infrastructure layer is heading — regardless of whether Ghost specifically wins the category.

The screenless box is a strange object. It's also a coherent bet on a world where your AI agents work harder than you do.

---

Is your company eligible? Submit to the leaderboard → onepersonunicorn.co/submit