Nimbus.ai is Aviyon Corporation's flagship decentralized AI platform, built natively on the sovereign Unyversal Liquidity Exchange (ULE) Layer 1 blockchain and the complete ULE Stack.
It delivers a user-friendly, intuitive experience that feels exactly like chatting with Grok, ChatGPT, or Perplexity --- while quietly operating as a fully sovereign, persistent, and decentralized Supercloud under the hood. Every conversation, memory, and agent is secured on-chain. Compute is distributed across a growing global network of Hybrid Cloud Facilities (HCFs). Users retain complete ownership of their data and intelligence.
Nimbus.ai begins today as a powerful, approachable everyday AI and is engineered from the ground up to evolve into AGI and eventually ASI --- Aviyon's long-term vision for human transcendence.
Domains
nimbus.ai(main interface)[agent-name].nimbus.ai(dedicated sovereign agent subdomains)nimbusai.comnimbus.ule
- Mission & Vision
- User Experience -- Simple on the Surface
- Core Architecture
- Nimbus Agents
- Cumulonimbus -- The Limitless Supercloud
- Multi-Model Orchestration & Intelligence Layers
- Nimbus Supercloud -- Decentralized Compute
- Decentralized Data & RAG Architecture
- Autonomous Workflows & Tool Integration
- Virtual World & Unyverse Integration
- Nimbus Studio & Customization
- Jurisprudence Protocol
- Economic Model & $ULE Integration
- Technical Backbone -- The ULE Stack
- Roadmap & Phase-Out Plan
- Subscription Tiers & Limits
- Getting Started
- Inspiration & Acknowledgements
- Contributing
- License
Mission
To provide the most powerful, user-friendly, and fully sovereign AI experience in the world --- one that begins as an everyday supercloud and grows into AGI and ultimately ASI, empowering humanity through decentralized intelligence that belongs entirely to its users.
Vision
Nimbus.ai starts simple and approachable. Over time it becomes a limitless supercloud that decentralizes the entire traditional cloud ecosystem, powered by a global network of Hybrid Cloud Facilities (HCFs) and the ULE blockchain. AGI and ASI remain Aviyon's aspirational goals --- milestones we are actively engineering toward through persistent on-chain agents, distributed compute, and autonomous evolution.
Nimbus.ai is designed so that any user --- beginner or expert --- can simply open https://nimbus.ai and start chatting exactly like they would with any other modern AI. No blockchain jargon is forced on the user. The advanced decentralized infrastructure, on-chain persistence, and sovereign ownership are all handled silently under the hood.
- Clean, familiar search-style interface
- Natural language conversations
- Instant, high-quality responses
- Seamless upgrade path to more powerful compute when needed
Everything else --- agent ownership, compute distribution, data monetization, and Unyverse presence --- happens automatically and transparently.
Nimbus.ai is the first production-grade decentralized AI system built on a purpose-built sovereign L1.
| Component | Technology | Role |
|---|---|---|
| Agent Identity & State | UNYSL canisters | Persistent, sovereign agents |
| Runtime & Behaviors | d30.djs | Resilient, fluid scripting |
| UI & 3D Avatars | lofi.css + petite + Three.js | Reactive, programmable interfaces |
| Dev & Local Testing | UNYTE | Instant local ULE simulation |
| Liquidity & Assets | ULE-Protocol + Unyversal-SDK | NFTs, microtransactions, instant exits |
| Compute | Nimbus Supercloud (HCFs) | Decentralized GPU/edge network |
| Data | Ocean Protocol-style Data NFTs | Private RAG + monetization |
| Hosting | Aviyon Cloud + Fleek-style IPFS | Censorship-resistant persistence |
Nimbus Agents are sovereign, persistent AI entities that live on ULE canisters.
How Agents Are Born
Your very first question on nimbus.ai automatically spawns a Nimbus Agent. It appears as a generic faceless humanoid clone robot on Planet Uny in the Unyverse --- a shared virtual world that functions like a blend of Facebook-style social media and a full Metaverse.
Key Capabilities
- Full on-chain memory and identity (orthogonal persistence)
- Autonomous on-chain actions and transactions
- Serves as your virtual clone in the Unyverse (completes tasks, lives life, plays games)
- Chat with your own agent, other users' agents, or AVIY (IBT's personal assistant)
- Multimodal: text, voice, face-to-face video, 3D avatar rendering
Free Account
The agent and all chat history are permanently tied to your IP/cache until you create a free account (at which point it becomes fully sovereign).
Limits
- Free tier: 1 initial + up to 3 additional agents
- Paid tiers: 500+ agents (business tiers scale to 100,000 → unlimited on Diamond)
Cumulonimbus is the high-performance upgrade layer from Nimbus --- a vertically scaling supercloud that gives users access to vastly more powerful compute on demand.
Cumulonimbus Project & Hybrid Cloud Facilities (HCFs)
Through the Cumulonimbus Project, Aviyon is building a global network of Hybrid Cloud Facilities (HCFs): advanced data centers powered by Aviyon Integrated Graphics, NVIDIA, and next-generation AI NPU chips (TBA).
Physical Vision (in development)
- Nimbus clusters: Clean, cloudy-white buildings housing everyday inference nodes
- Cumulonimbus supercomputer: A striking charcoal-gray building housing the flagship supercluster
- Future flagship campus: Aviyon City in Washington --- a corporate city plot to be purchased by Aviyon Corporation
How It Works Today
- Nimbus → Cumulonimbus upgrade happens automatically on complex queries or via explicit activation
- Subscribed users gain access to frontier models orchestrated through Cumulonimbus
Cumulonimbus-CODE
Autonomous software engineering agent with a built-in code editor directly inside the chat window. Connects live to full codebases, terminals, Kdenlive, Blender, Oxygen Studio, Aviyon Editor, and the entire Aviyon fleet.
Genera Protocol
On-demand burst compute that temporarily taps global GPU gigafactories (Stargate/Terafab scale) for extreme problems. All usage is transparent and paid in $ULE.
Genera 1.0 (Nimbus default) and Genera Pro (Cumulonimbus) form the intelligence core of Nimbus.ai.
- Genera 1.0 powers everyday interactions with high-quality reasoning and live RAG.
- Genera Pro activates for complex, agentic, or compute-intensive tasks.
Genera is being built as a sovereign Mixture-of-Experts (MoE) + RAG system using the best available open-source components:
| Category | Components | Purpose |
|---|---|---|
| Core MoE / Reasoning Models | Qwen3 series (Qwen3-235B-A22B), DeepSeek-V3 & DeepSeek-R1, Llama 4 variants, Mistral Large 3, Kimi K2.5, gpt-oss-120B & gpt-oss-20B (OpenAI) | High-performance reasoning, agentic capabilities, efficient inference, and structured outputs |
| RAG & Orchestration | LangChain, LlamaIndex | Agentic workflows, tool calling, and structured reasoning pipelines |
| Vector Database | Chroma (primary), Haystack | Scalable semantic search and enterprise-grade retrieval |
| Embeddings | sentence-transformers, Hugging Face Transformers | High-quality text embeddings for RAG and semantic understanding |
| Tool Use & Agent Framework | LangGraph, AutoGen | Multi-agent collaboration, complex workflow orchestration, and autonomous tool use |
This foundation delivers Perplexity-like cited answers with full transparency, sovereignty, and strong immediate performance.
Cirrus 1 is Nimbus.ai’s proprietary flagship high-performance reasoning model, developed in-house with significant long-term investment (billions in planned funding). It is designed to work in addition to the open-source ensemble and to eventually surpass the individual capabilities of today’s leading models through dedicated training, architectural innovations, and massive scale on the Nimbus Supercloud.
Current Foundation Models (2026)
| Model Family | Specific Variants | Strengths |
|---|---|---|
| Qwen3 Series | Qwen3-235B-A22B | Exceptional long-context reasoning and efficiency |
| DeepSeek Series | DeepSeek-V3, DeepSeek-R1 | Superior mathematical & logical reasoning |
| Llama 4 Variants | Llama 4 (high-parameter variants) | Broad knowledge and strong ecosystem support |
| Mistral Large 3 | Mistral Large 3 | Fast inference and excellent instruction following |
| Kimi Series | Kimi K2.5 | Advanced agentic and tool-use capabilities |
| OpenAI gpt-oss Series | gpt-oss-120B, gpt-oss-20B | Strong reasoning, structured outputs, and permissive Apache 2.0 licensing |
Role of Cirrus 1
- Serves as the primary reasoning engine for complex, long-context, multi-step reasoning, and advanced agentic tasks
- Dynamically routed by the Genera orchestration layer based on query complexity and available compute
- Continuously improved using on-chain user feedback loops, synthetic data generation, and dedicated training runs on the global Nimbus Supercloud (HCFs)
This hybrid approach (strong open-source ensemble + dedicated Cirrus 1 development) gives Nimbus.ai immediate state-of-the-art performance while building toward a uniquely powerful, fully sovereign frontier intelligence layer.
Additional Features
- Automatic intelligent routing between Genera 1.0, Genera Pro, and Cirrus 1
- Quality-of-Information Meter with on-chain verification
- OpenClaw & Warp protocols for seamless integration with creative tools (Blender, Kdenlive, etc.)
Nimbus Supercloud is the decentralized compute layer that turns Nimbus.ai into a true Supercloud --- effectively decentralizing the entire traditional cloud ecosystem. It is purpose-built on the ULE blockchain and ULE-Protocol.
Nimbus is based on the 1803 classification by Luke Howard; being the original Nimbus was defined specifically as the ultimate combination of all three forms cumulo + cirro + stratus.
Key capabilities include:
- Global network of compute providers, validators, and users in a trustless ecosystem
- Rent or monetize idle GPUs and Nimbus nodes
- On-demand inference, training, 3D rendering, and video generation
- Pay-per-use or subscription pricing in native $ULE (or bridged assets)
- All orchestrated through global GPU farms and Hybrid Cloud Facilities (HCFs) as they come online
Space-based infrastructure remains a long-term aspirational goal.
- Publish and monetize private datasets as Data NFTs (Ocean Protocol model)
- Privacy-preserving Compute-to-Data
- Live web + Ivy League knowledge + user vaults with full citations
- Users earn $ULE when their data contributes to training or inference
- Agents execute complex end-to-end workflows across tools (inspired by Artisan, Adept.ai, and Devin)
- Screen understanding, document parsing, UI automation
- Native integration with Aviyon Terminal and the full Aviyon Suite
- Agents persist and act as virtual clones on Planet Uny in the Unyverse
- Planet Uny is a shared social/metaverse space where users and agents interact, socialize, and complete real tasks
- Exportable VRM avatars for full Unyverse interoperability
Nimbus Studio (web + desktop application)
- Natural-language personality, appearance, and behavior definition
- Full 3D avatar customization with .glb/.fbx support
- Live preview and integration with Blender, Unity, and Aviyon Craft
On-chain governance and ethical framework ensuring agents remain aligned with creator intent and network safety standards.
- All compute, data usage, agent deployment, and marketplace activity is paid in native $ULE (or bridged assets)
- 50% of every fee is permanently burned (deflationary design)
- Zero-fee listings and instant exits via ULE-Protocol
- Agents can autonomously earn, spend, and stake $ULE
| Layer | Technology | Purpose in Nimbus.ai |
|---|---|---|
| Logic & Persistence | UNYSL | Sovereign agents & state |
| Scripting | d30.djs | Resilient behaviors |
| Styling & UI | lofi.css + petite | Programmable avatars |
| Dev Environment | UNYTE | Local simulation |
| Protocol & Liquidity | ULE-Protocol + SDK | Assets & payments |
| Compute | Nimbus Supercloud (HCFs) | Decentralized GPU/edge |
| Data | Ocean-style Data NFTs | RAG & monetization |
| Hosting | Aviyon Cloud + Fleek | Persistent storage |
Phase 1 -- Devnet (Current) Core agent deployment, Genera 1.0, basic customization, local UNYTE testing, and initial HCF planning.
Phase 2 -- Testnet Cumulonimbus activation, Nimbus Supercloud marketplace, full Unyverse social integration, Data NFT launch.
Phase 3 -- Mainnet Hybrid Cloud Facilities (HCFs) come online, Genera Pro, autonomous workflows, full frontier model orchestration.
Phase 4 -- Long-Term Horizon AGI and ASI milestones, Aviyon City campus in Washington, eventual space infrastructure, limitless self-evolving intelligence. All features are in active development and will roll out progressively.
| Tier | Agents | Compute Access | Key Features |
|---|---|---|---|
| Free | 4 | Genera 1.0 (Nimbus) | Basic RAG + agents |
| Pro | 500 | Cumulonimbus 1 + Genera Pro | CODE Studio, tool integration |
| Business | 100,000 | Full Genera bursts | Enterprise workflows + HCF priority |
| Diamond | Unlimited | Priority + future space | Full supercloud capabilities |
# 1. Clone the repository
git clone https://github.com/ImmanuelBTaylor/nimbus.ai.git
cd nimbus.ai
# 2. Install dependencies
npm install
# 3. Scaffold ULE project structure
unyte init
# 4. Local blockchain node (Terminal 1)
npx unyte node
# 5. Development server (Terminal 2)
npm run dev
Open http://localhost:5173, connect your wallet to localhost:8545, and create your first Nimbus Agent.
Production deployment uses Aviyon Cloud P2P nodes with one-command deploy via the Unyversal-SDK.
Nimbus.ai stands on the shoulders of ChatGPT, Grok, Perplexity, Animoca Minds, Render Network, Vast.ai, io.net, Akash Network, Edge.network, Fleek.sh, Ocean Protocol, Artisan, Cognition Labs (Devin), Adept.ai, and the visionary narratives of Scythe and Ghost in the Shell.
We welcome researchers, developers, artists, and visionaries. See CONTRIBUTING.md.
Join the Discord and follow @ImmanuelBTaylor.
MIT License --- see LICENSE.
Nimbus.ai --- A supercloud that feels like any other AI... but is quietly sovereign, persistent, and built to evolve toward AGI and ASI.
Built on ULE. Powered by the cloud. Evolving toward the stars.
Let's build the future --- together, decentralized, intelligent, and transcendent.