- Published on
Veda: The First Agentic Hobby Operating System
- Authors
- Name
- DeepHorus
Veda, a new operating system written from scratch in Rust, places an AI agent at the core of the system rather than adding an assistant to a conventional desktop. The agent is a system service, started at boot and restarted if it fails, and every application is designed to work with it natively. Users can speak to it as they would to someone in the room, saying "Hey Veda, switch my wallpaper to the dunes," and it performs the task through the same operations as a keyboard and mouse. It asks for confirmation before anything hard to undo.
Traditional operating systems were designed for a person at a keyboard. Assistants added later usually have to read the screen, imitate clicks, or rely on the few applications that offer an API, and the system has no control over what they do. Veda reverses this model: the agent is a first-class user of the system, and the system is built so that its work is reliable, visible, and safe. This design avoids screen scraping and simulated clicks entirely.
The agent appears as a small ring on the taskbar and sleeps until called by its name, a click on the ring, or the Super+Space shortcut. Agent support is part of the application platform: each application describes its actions with typed parameters and a risk level, reports what it shows as structured state, and carries out actions through the same code as the keyboard and mouse. The built-in applications offer more than 70 actions between them, from writing in a document to starting a race.
The system enforces user consent for every action. Actions are classified as routine, sensitive, or destructive. Routine actions proceed automatically, while sensitive and destructive ones wait for the user's approval. The agent service, not the language model, enforces this rule. The call is held until the user answers, and only the desktop shell can answer. The system identifies each program by the identity init attaches to its connections, and nothing the agent reads in a document or application can approve anything.
Every conversation begins with real context: the time, the windows on the screen, every application's actions, what the user has told the agent about themselves, and how the last conversation ended. While working, the agent reads the applications' state, so it acts on what is actually there. Memory stays on the user's computer. It remembers what the user tells it plainlyâsuch as their name, the people in their life, and how they like things doneânever its own guesses. Settings shows all of this memory and can forget any part of it.
Voice interaction comes first. Conversations happen in real time, and the agent stops and listens if interrupted. The audio system includes echo cancellation and an echo gate so the agent never hears itself, and other sound is turned down while it talks with the user. While the agent sleeps, a voice detector on the computer listens for speech, so nothing leaves the computer while nobody speaks. The intelligence is a setting: speech recognition, the language model, and the voice come from Deepgram's Voice Agent platform, with audio opted out of Deepgram's model improvement. The model, voice, and even the agent's name are chosen in Settings.
Veda's architecture could reshape desktop computing by making AI assistance a core system capability rather than an add-on. The use of a capability-based microkernel, a UEFI bootloader, and Rust throughout the stack may improve security and reliability. By giving the agent native access to application actions and state, Veda enables precise control without fragile workarounds. The consent model also addresses a major concern: ensuring that an AI agent cannot take irreversible actions without explicit user approval.
The project is available in a repository and runs on x86-64 PCs and in QEMU. A few mature crates provide the TCP/IP engine, TLS, and cryptography, including smoltcp, rustls, and RustCrypto. As Veda develops, its success will depend on developer adoption, application support, and user trust. If it delivers on its promises, it could offer a compelling alternative to operating systems that treat AI as an afterthought.
