HyperionOS – an intent-native operating system HyperionOS, described as the first intent-native operating system, manages user goals directly instead of processes, files, or applications. Built with a capability-secured kernel and a unified scheduler balancing compute, inference tokens, and context windows, it uses an Intent Engine, Context Engine, and Model Router to turn language into sub-goals and select appropriate models. The system replaces traditional folders with a knowledge graph of Semantic Objects and orchestrates multi-agent workflows with logged decisions. You ask. It understands. Hyperion is the first intent-native operating system. Philosophy Your goals are the operating system. Every prior operating system asked you to manage proxies for what you wanted: a process, a file, a window, an application. Hyperion manages what you actually have. - Processes - Goals - Threads - Intentions - Files - Knowledge - Windows - Context - Applications - Memory, reasoning, capabilities Why Hyperion You shouldn't think like a computer. - Finding files - Choosing applications - Remembering workflows - Managing folders - Installing software - Complex settings Hyperion clears all of that away and gets to work. Architecture You speak. Everything else listens. Hyperion is built in layers, each one building on the one below it. Closer to the hardware, it's fast and safe. Closer to you, it understands what you mean. Kernel · Where it begins The hardware layer everything else stands on. It's open source, so its safety claims are checkable. Kernel · Where it begins The hardware layer everything else stands on. It's open source, so its safety claims are checkable. Capability-secured from the ground up: nothing crosses a trust boundary without an explicit, revocable grant. System runtime · How it stays fast and safe Schedules work across CPU, GPU, memory, and battery, keeping every process in its own boundary. System runtime · How it stays fast and safe Schedules work across CPU, GPU, memory, and battery, keeping every process in its own boundary. A unified scheduler balances compute, inference tokens, and context windows, the way earlier operating systems balanced CPU and RAM. Platform services · What it's built from Reusable capabilities, storage, updates, and networking. Platform services · What it's built from Reusable capabilities, storage, updates, and networking. Capabilities are Hyperion's replacement for the application: a declared contract, a trust level, and one or more interchangeable implementations. Knowledge · What it knows and remembers Your information, connected by meaning. Knowledge · What it knows and remembers Your information, connected by meaning. Every document, photo, message, or project is a Semantic Object with typed relationships to everything else: a knowledge graph standing in for folders and filenames. Cognition · Where it thinks Understands what you're asking for, recalls what's relevant, and picks the right model for the job. Cognition · Where it thinks Understands what you're asking for, recalls what's relevant, and picks the right model for the job. The Intent Engine turns language into a graph of sub-goals. The Context Engine attaches what already exists a calendar event, a past conversation . The Model Router picks local or cloud reasoning per task. Coordination · How the work gets organized When a goal needs more than one specialist, this layer decides who does what, and keeps a record of why. Coordination · How the work gets organized When a goal needs more than one specialist, this layer decides who does what, and keeps a record of why. Multi-agent orchestration assigns sub-goals to specialized agents and resolves conflicts between them. Every decision is logged for you to inspect. Experience · What you see and say Conversation, generated screens, and voice: the only parts of Hyperion you directly touch. Experience · What you see and say Conversation, generated screens, and voice: the only parts of Hyperion you directly touch. The Dynamic UI Runtime assembles a Workspace on demand for whatever you're doing, then tears it down when you're done. Accessibility is built into that runtime. How Hyperion works You only need one sentence. You never pick an application. You say what you need, and Hyperion figures out the rest, visibly and explainably. “Help me prepare for tomorrow's interview.” Reads your calendar, your resume, and past notes. You don't repeat yourself. - Company research - Practice questions - Flashcards - A timer - The job posting “I need to launch my startup.” Breaks one goal into a plan, and starts on what's ready first. - Market research - Business plan - Branding - Entity formation - A website - Launch checklist “Create a social media graphic from these photos.” Instead of: open Photoshop, open Finder, export PNG, save as. - Photos selected - Layout composed - Export ready Live console You can watch it run. Every session below is copied from an actual hyperion-console run. Nothing here was written for the page. A new user's first session Hyperion remembers what you told it a moment ago, breaks one goal into several tasks, and can hand back any single result by name. my name is Alex✓ status: done -- mock model 1 echo: my name is Alex what is my name✓ status: done -- mock model 1 echo: Recent conversation, most recent last: my name is Alex Now respond to: what is my name launch my startup✓ market research: Done business model: Done branding: Done legal formation: Done /result market research"market research"'s real result, recorded just now: mock model 1 echo: Provide a concise research summary about market research -- launch my startup. ... scenarios/onboarding-new-user-launches-startup.txt https://github.com/JGalego/HyperionOS/blob/main/scenarios/onboarding-new-user-launches-startup.txt Watching the knowledge graph grow One command shows what Hyperion has learned, before and after it acts, as a graph you can inspect yourself. /graphThe knowledge graph is empty -- nothing recorded yet. launch my startup✓ market research: Done business model: Done branding: Done legal formation: Done /graph9 nodes: 0 intent -- you asked: "launch my startup" 90% confident 1 intent -- a planned task: market research 2 intent -- a planned task: business model 3 intent -- a planned task: branding 4 intent -- a planned task: legal formation 8 task result -- a result: mock model 1 echo: Provide a concise research summary... 7 edges: 5 2 --depends on-- 1 weight 1.00 9 1 --produced-- 8 weight 1.00 scenarios/knowledge-graph-growth-audit.txt https://github.com/JGalego/HyperionOS/blob/main/scenarios/knowledge-graph-growth-audit.txt One question, three models OpenAI, Anthropic, and Groq each answer the same question in the same session, switched live with no restart. /backend openai gpt-4o-mini✓ Switched to the openai model "gpt-4o-mini" backend. what is the smallest country in the world by land area✓ status: done -- The smallest country in the world by land area is Vatican City. It has an area of approximately 44 hectares 110 acres or about 0.17 square miles... /backend anthropic claude-haiku-4-5-20251001✓ Switched to the anthropic model "claude-haiku-4-5-20251001" backend. what is the smallest country in the world by land area✓ status: done -- The Smallest Country in the World by Land Area Vatican City is the smallest country in the world by land area, with only about 0.44 square kilometers ... /backend groq llama-3.3-70b-versatile✓ Switched to the groq model "llama-3.3-70b-versatile" backend. what is the smallest country in the world by land area✓ status: done -- The smallest country in the world by land area is the Vatican City, with an area of approximately 0.44 km² 0.17 sq mi ... scenarios/cloud-provider-comparison.txt https://github.com/JGalego/HyperionOS/blob/main/scenarios/cloud-provider-comparison.txt A node that can't translate, asks one that can Six real, separately-launched hyperion-console processes discover each other over real mDNS, each backed by a different real cloud model Anthropic, OpenAI, Groq -- no human ever names a host, port, or peer. /backend groq llama-3.1-8b-instant✓ Switched to the groq model "llama-3.1-8b-instant" backend. /mesh-request 9106 translate-ja say hello in JapaneseI don't have "translate-ja" myself, so I asked kenji. hyperion-a2a. tcp.local. 127.0.0.1:9101 , which does. It replied: status: done -- こんにちは Konnichiwa This is a common way to say "hello" in Japanese, used during daytime hours. Other greetings include: - おはよう Ohayou - good morning - こんばんは Konbanwa - good evening Trusting 127.0.0.1:9101's identity for the first time: 46c559994406105160c06ed3e3182cbe229d2dd27dee297afc2fbf70745dc0d1. scripts/run-mesh-demo.sh https://github.com/JGalego/HyperionOS/blob/main/scripts/run-mesh-demo.sh Features You will notice the difference. You never have to remember where you saved something. Everything is connected by meaning, so you can always find it. Hyperion assembles the right tools automatically. Capabilities are selected and composed for the goal. You never choose an app. It can always tell you why. Why, how, with what evidence, at what confidence, and how to undo it, on demand. Your data stays on your device, by default. The cloud is an explicit, consented upgrade. It grows as you do. Beginners see plain language and guided steps; experts get the shell, the API, the shortcuts. You can always undo it. Every action is interruptible, and every change can be rolled back. Getting started You can read every line of it. Grab a signed, ready-to-flash release image, or build every layer yourself from source. - 1 Download the image Every tagged release publishes a ready-to-flash x86 64 disk image on GitHub Releases. https://github.com/JGalego/HyperionOS/releases/latest hyperion-x86 64-