• Ollama updated to the latest Nix version (0.12.2)
• Open WebUI updated to the latest Nix version (0.6.31)

Check out the code on the GitHub or try the new version on OpenxAI Studio

• Xnode Auth v1.1.0 released to main branch
  • Access control can now be defined in external sources
    • OS set permission not needed anymore to update auth
    • Easier for custom applications to control (partial) auth
  • Filesystem and http(s) external source fetching support added
  • External sources should be restricted to mitigate external source breach from propagating
    • External source can be restricted to only control auth on select domains
    • Per source per domain rules on allowed users and paths
  • External sources can be updated through restarting the associated xnode-auth-source-update systemd unit
    • Allows for sub-second authentication updates, instead of up to a minute system rebuilds

• Xnode Manager v1.0.1 released to main branch
  • Added filesystem ACL get and set endpoints (/file/{scope}/get_permissions and /file/{scope}/set_permissions)
    • Allows for more granular file permission control
  • Added users and groups endpoint (/info/users/{scope}/users and /info/users/{scope}/groups)

• Progress Update
• This week saw record hackathon engagement, major Studio upgrades, and stronger recognition of tokenized GPUs as a cornerstone of permissionless AI.
• Ecosystem Growth
  • India: Hacknode India became the largest Web3 hackathon to date, with 2,501 registrations, 343 submissions, and 388 pull requests. Workshops and bootcamps fueled momentum.
  • Australia & Global: Hacknodes were extended to increase participation, and the Hacknode Dashboard went live with links, submissions, and resources.
  • Offchain: Outreach expanded across 40 country channels. Local events and ambassadorships were prepared, with Luxembourg confirmed for October and strong interest from Rome, Turkey, and London.
• Engineering & Product
  • Studio v4.5: A new landing page with live stats was launched, demo servers scaled to 10, and responsiveness improved by disabling auto-gen in the chat app. Promo codes were integrated for GPU credits.
  • Studio v5: Development advanced with new transitions, homepage redesign, and improved responsiveness. Optimized animations were being prepared.
  • Base Mini App: The first Nixified demo app was deployed, showing how developers could host through the base app store.
• Narrative & Positioning
  • OpenxAI framed itself as defining permissionless AI. Tokenized GPUs (tGPUs) emerged as a leading application at the AI-DeFi intersection, enabling staking, fractionalization, and programmable GPU access.
• Community & Engagement
  • Ambassadors produced content simplifying GPU credits.
  • Micro-KOLs were activated for hacknode promotion.
  • A gamified Discord Academy was put in motion.

• Setting added to disable auto generation to improve performance and responsiveness
  • This is enabled by default when deploying onto Demo Servers to improve the user experience when trying out slower models

Check out the code on the GitHub or try the new version on OpenxAI Studio

• OpenxAI Demo Servers increased from 3 to 10

Deploy any app on one of the new demo servers on OpenxAI Studio

• Hack Node Boot Camp: Your All-in-One Guide to Hack Node Success
  The “Hack Node Boot Camp: Your All-in-One Guide to Hack Node Success” was a comprehensive workshop hosted by OpenxAI Events on Google Meet. The event aimed to equip participants with the skills and knowledge needed to excel in the Hack Node initiative and beyond.
• Key Highlights of the Workshop:
  • Introduction to OpenxAI: Participants gained insights into the mission and vision behind OpenxAI, the first permissionless AI protocol.
  • Opportunities with Hack Node: Attendees learned about exciting job opportunities and how participation in Hack Node could serve as a gateway to the Global Accelerator program.
  • Vibe Coding: The workshop covered the latest coding approach designed to help developers build faster and smarter applications.
  • Building for Xnode: Developers explored how to integrate Ollama open-source models and Postgres databases into their projects, enhancing their development capabilities.
  • Base Miniapps: Attendees understood why Base provides an ideal platform for creating scalable, decentralized AI applications.
  
  The boot camp offered a valuable opportunity for developers and AI enthusiasts to deepen their understanding of decentralized AI development and to connect with the OpenxAI community.

• Progress Update
• Ecosystem Growth
  • India: Hacknode India became one of the largest Web3 hackathons with 1,500+ registrations, 34 deployments, 324 demo apps, and 101 publications, ranking among the top 10 initiatives for developer adoption.
  • Australia: Hacknode Australia reached 124 applicants and 9 project submissions, with developers advancing “Nixification” to prepare projects for scalable deployments.
  • Global: Hacknode Global launched across multiple regions, expanded partnerships including Offchain’s 80+ chapters, and prepared for physical events starting in September.
• Engineering & Product
  • Released Version 4.5 with deployment tracking, promo codes for GPU trials, automated vesting schedules, and a new changelog site.
  • Began V5 development with multi-chain features and tools for benchmarking deployment costs and efficiency.
• Community Distribution
  • Discord activity grew daily through campaigns and hacknode participation.
  • 129 micro-KOLs promoted Hacknode with a new reporting system to track performance.
  • The first community AMA was scheduled for August 25, featuring Ashton Hettiarachchi and Santeri Aramo.
  • Outreach to university groups continued, with focus shifting toward Web3-native communities for adoption and visibility.
• Narrative & Differentiation
  • We clarified OpenxAI’s unique position as OwnAI, a permissionless AI model combining decentralization with sovereignty.
  • Each milestone, deployment, and partnership reinforced this first-mover advantage.

• Hack Node Developer Workshop: Vibe Code Your First AI Mini App
  On August 26, 2025, OpenxAI hosted the "Hack Node Developer Workshop: Vibe Code Your First AI Mini App" via Google Meet, led by Developer Relations lead John Forfar. The session aimed to empower participants to build decentralized AI applications using OpenxAI's ecosystem.
• Workshop Highlights
  • Vibe Coding Setup: Attendees learned to configure their local development environments for "vibe coding," a streamlined approach to building decentralized applications.
  • Base Integration: Participants were guided through the deployment of their applications on Base, OpenxAI's decentralized platform, ensuring seamless integration with the ecosystem.
  • Ollama LLM Backend: The workshop demonstrated how to integrate Ollama as the AI model backend, providing participants with access to powerful language models for their applications.
  • Data Management with Postgres: Attendees explored how to collect and manage data using a Postgres database, facilitating efficient data handling within their AI applications.
  
  The event was designed for both beginners and those looking to enhance their skills in decentralized AI development. It provided practical experience and confidence in building AI-powered applications within the OpenxAI ecosystem.

Cyberscope audit completed without any issues for the following contracts:

• OpenxAI (OPENX): ERC20 Governance Token
• OpenxAI Claimer: to claim off-chain rewards on-chain
• OpenxAI Genesis: tracking of asset transfer contributions on a first-come first-serve basis

The full audit report can be found on the GitHub

• Progress Update
• Marketing & Communications
  • Media Strategy: Narratives were expanded to target builders, researchers, and the broader ecosystem, ensuring OpenxAI’s story reached each audience in the right context.
  • Content Calendar: A structured content calendar anchored communications, aligning product milestones with community highlights for consistent visibility.
  • Formats: Standardized formats for AMAs, announcements, and long-form content enabled faster publishing without compromising quality.
  • User Guides: Practical onboarding guides were developed to help new users begin interacting with the network seamlessly.
• Ecosystem Growth
  • Australia: Hackathon activity picked up with 15 pull requests, 7 project submissions, expanded university outreach, and the first Sydney event.
  • India: University events and a dev workshop engaged 500+ students, amplified online by 800+ posts that drove 2,000+ website visits.
  • Global: Hacknode Global launched with 20 participants, early university outreach in Europe, and new discussions with hardware and media partners.
• Engineering
  • Version 4.5 was completed, delivering major upgrades ahead of Version 5:
  • Added deployment tracking across versions for transparent model monitoring.
  • Enabled decentralized compute servers for persistent private environments.
  • Automated access rights and resource allocation to reduce developer friction.
  • Integrated the open-source ChatGPT model into OpenxAI Studio for easy deployment.
  • Refined the user experience with new animations and faster model management.
• Community
  • Surpassed 4,100+ active members on Discord.
  • Welcomed ambassadors with 10,000+ followers amplifying OpenxAI across X and other platforms.
  • Saw students and developers in India publish independent blogs and articles, boosting organic visibility.
  • Generated hundreds of LinkedIn and Twitter mentions from community-led discussions.
  • Implemented standardized reply sheets and reposting workflows to spotlight community content.
• Content & Media
  • Transformed long-form hackathon, AMA, and workshop recordings into short clips for YouTube, TikTok, and X.
  • Rolled out new intro/outro templates for consistent OpenxAI branding across all media.
  • Revived the monthly newsletter for 5,000+ subscribers with key updates and stories.
  • Mirrored YouTube content systematically across X, LinkedIn, and Discord for wider reach.

• Ollama updated to the latest Nix version (0.11.4)
  • gpt-oss is now supported
• CUDA accelerated Ollama example config added and cached

Check out the code on the GitHub or try the new version on OpenxAI Studio

• Progress Update
• System Readiness & Product Coordination
  • Core smart contracts and front-end systems were completed and aligned, with efforts focused on calibration, testing, and content coordination under a community-first strategy.
  • The “Own AI” narrative gained traction as a central theme across communications and ecosystem storytelling.
• Adoption & Community Growth: Real-World Adoption Begins
  • India Hack Node: OpenxAI Studio saw real use in India, with 60+ participants deploying models, 4 university MOUs signed, and campaigns underway including an upcoming Financial Express article.
  • Australia Hack Node: Over 170 job applicants and 50 hackathon sign-ups joined, with mentoring sessions shared via YouTube and projects ranging from AI games to sustainability apps.
  • United States Hack Node: The US track launched through 30+ organic traffic sources without paid promotion, reflecting steady, community-led growth.
  • Global Hack Node: A global track was developed to meet demand from Europe, Nigeria, and beyond, with rollout expected in the coming weeks.
  • Philippines Events: Offline event planning began with BASE Protocol ambassadors, exploring Farcaster-integrated mini-apps and community collaborations.
• Studio Progress: v4.5 Implementation and v5 Prototypes
  • V4.5: The deployment interface was implemented with improved usability, mobile responsiveness was finalized, and prototypes were temporarily hosted for feedback.
  • V5: Feature planning introduced multi-chain settings starting with Base, alongside a modular app store and navigation upgrades, with a Dubai event prepared to showcase the vision and live OpenxAI agents.
• Community Activation Strategy
  • A structured activation plan emphasized conversion and contribution through trusted networks, creator collaborations, regional ambassadors, and AI-native expansion, supported by a live contribution tracker for transparency.
• Compute Scaling & Multichain Functionality
  • Per-application GPU access was supported, tokenized GPU credits were planned, and negotiations for long-term GPU discounts were initiated.
  • Deployments were structured across Base, BNB, Solana, and Bitcoin, with Studio apps prepared to support ecosystem-specific logic and custom deployment paths, marking a shift in how intelligence is created, owned, and shared.

• Xnode Manager now allows per container access control to NVIDIA GPUs
• Specify the nvidia_gpus array in set container settings with the ids (/dev/nvidia{id}) of the NVIDIA GPUs you'd like to be accessible inside this container

• Discord Dev Workshop: Code Like a Pro with OpenxAI Tools On August 6, 2025, OpenxAI conducted a live coding session hosted by John Forfar and led by Lead Engineer Samuel Mens. The workshop focused on developing custom applications using OpenxAI tools, featuring:
• Live Coding Demo: Demonstrating the integration of OpenxAI models into applications.
• Q&A Session: Addressing participant questions and providing guidance.
• Hack Node Australia: Encouraging students to build AI, Web3, or gaming applications for the Hack Node Australia event.

• XnodeOS can now be deployed automatically on Hyperstack using Xnode Deployer

• Progress Update
• Adoption & Community Growth
  • This week focused on expanding grassroots adoption across universities and developer ecosystems:
  • Australia Hack Node: Reached 7,500 students at the University of Sydney via official email; accredited for PEP hours with academic credit.
  • India Hack Node: Officially approved and set to launch August 10; registration and campaign video released.
  • USA Hack Node: Onboarded ambassadors, initiated outreach to top campuses, and prepared for the first US hackathon.
  • Career Funnel: Shifted Australia messaging to long-term career paths with job posts, orientation days, and Open R&D mini-projects.
  • Developer Bootcamps: Launched 5-day Core Bootcamp with demo apps and workshops feeding into hackathons.
  • University Events: Finalized events at PCTE, NIT Jalandar, and CT University; branded kits dispatched to chapters.
• Marketing Momentum
  • This week the team aligned narratives and prepared for upcoming visibility pushes:
  • Launched PR strategy, initiating talks with key media and crypto journalists.
  • Finalized external messaging for clear, consistent communication.
  • Rolled out a comprehensive media kit for PR and influencer outreach.
  • Defined strategy for engaging micro-creators and KOLs on X, TikTok, and YouTube.
  • Collected campaign specs for the next product release with creator partners.
• Engineering & Product
  • The team made major strides this week as OpenxAI scaled toward its next phase:
  • OPENX Claiming Contract: Tracked total vs. claimed amounts, fixed double-claim risks, and enabled governance features.
  • Frontend: Completed v4.5 Studio UI design; began v5 App Store build.
  • Xnode: Published custom app docs, reworked deployment for flexibility, and restricted socket access for security.
  • Infrastructure: Chose Hyperstack as the GPU server provider to power workloads.

The following contracts have been updated:

OpenxAI (OPENX)

• Added ERC20 Votes (on-chain governance)

OpenxAI Claimer

• Keeps track of total amount of tokens claimed instead of per-proof id
  • Reduces the possibility of double spending attacks
  • Reduces the complexity for the off-chain indexer
  • Improves the user experience of not having to deal with past un-redeemed claims

OpenxAI Genesis

• Contributions are sent directly to the project treasury

You can check out the updated contracts yourself on the GitHub

• Xnode Manager now listens on a unix socket (/var/lib/xnode-manager/socket) instead of a TCP port
• Unix socket allows for stricter access control restrictions to reduce the chance of accidental exposure

• OpenxAI Global Accelerator Program: Orientation Workshop
  On July 19, 2025, OpenxAI hosted the Global Accelerator Program: Orientation Workshop, a pivotal event aimed at equipping developers with the tools and knowledge to succeed in the upcoming 30-day decentralized AI hackathon. The workshop, conducted via Google Meet, attracted 92 participants, including Mujeeb Nouman and Himagirisai Maredla.
• Workshop Highlights
  • Comprehensive Hackathon Overview: Attendees received step-by-step guidance on how to participate in the Global Accelerator Program, including instructions on signing up, selecting from nine project tracks, and forming teams.
  • Project Development Strategies: The session provided insights into brainstorming ideas, developing applications, and leveraging available resources, catering to participants with varying levels of experience in AI.
• Submission Process and Judging Criteria: Detailed information was shared regarding the submission process, deadlines, and judging criteria to help participants maximize their chances of success.
• Prizes and Opportunities: The workshop highlighted the $5,000 prize pool, which includes $5,000 in cash and an Xnode One Hardware valued at $3,500 USD, along with mentorship from NVIDIA and OpenxAI experts.
• Interactive Q&A Session: Participants had the opportunity to engage directly with the OpenxAI team, ask questions, and connect with fellow participants, fostering a collaborative environment.
  This orientation workshop marked a significant step in preparing developers for the Global Accelerator Program, emphasizing OpenxAI's commitment to fostering a decentralized, permissionless AI ecosystem.

• Xnode Manager v1.0.0 released to main branch

• Xnode Demo library updated to be compatible with the latest Xnode Manager version
• Now using Xnode Controller to manage user reservations
• Direct Xnode Manager access through Xnode Auth instead of request rewrites

• NextJS + Ollama template released for XnodeOS
• Allows websites using local AI based on this template to deploy seamlessly on OpenxAI Studio

You can use the new template on GitHub

• OpenxAI × NVIDIA Developer AMA
• OpenxAI hosted an AMA with NVIDIA to discuss decentralized AI infrastructure, GenAI, and developer sovereignty.
• Highlights:
  • The session featured NVIDIA’s Hariharan Suresh alongside OpenxAI Founder Ashton Hettiarachi and DevRel John Forfar, covering AI deployment at the edge and permissionless compute.
  • OpenxAI Studio was showcased, demonstrating how developers can train, deploy, and host AI models using $OPENX GPU credits and create tokenized AI apps on-chain.
  • Attendees included ML researchers, infrastructure providers, and Web3 founders exploring real-world AI deployments.
  • Studio v4 is set to include support for NVIDIA’s NIM, TensorRT, and NeMo, with new GPU nodes being onboarded to expand decentralized training and inference.

• Xnode Controller library released to allow projects to automatically update XnodeOS config based on external data
• The library allows you to define how to decide the current "controller" user and the XnodeOS configuration that will update when this "controller" changes

• Xnode Manager SDK library released for Rust

View on crates.io

• Xnode Manager on XnodeOS is now exposed on manager.xnode.local
• Uses a self-signed https certificate to be able to use encryption
• Requires a proxy (such as https://xnode.openmesh.network/api/xnode-forward) to access from the browser environment

• Removed built-in auth from Xnode Manager
• Default auth used by XnodeOS is the new Xnode Auth
• Xnode Auth allows for unified access control to all apps running on XnodeOS
• Xnode Manager can now also be used through different authentication layers, including custom implementations

• OpenxAI Developer Workshop On June 22, 2025, OpenxAI hosted a Developer Workshop aimed at empowering developers to build and deploy AI models on a decentralized, permissionless platform. The session, conducted via Google Meet, attracted over 130 attendees and was led by core contributors John Forfar and Akshay Kumar Sharma.
• Key Highlights:
• Introduction to OpenxAI Ecosystem: The workshop provided an overview of OpenxAI's suite of tools designed to facilitate decentralized AI development.
• Model Deployment Simplified: Participants learned how to build and deploy AI models on OpenxAI's platform, emphasizing ease of use and accessibility.
• Hands-On Experimentation: Attendees engaged in practical sessions, experimenting with AI models to gain firsthand experience.
• Showcase of "Mimic": The workshop introduced "Mimic," an OpenxAI offering that enables transparent and responsible AI usage through voice commands.
• Decentralization Vision: The session highlighted OpenxAI's commitment to decentralizing model ownership, data, and control, aligning with the broader movement towards open-source AI development.
• Attendee Benefits:
• Certificate of Participation: All attendees received certificates acknowledging their participation.
• Top Contributor Recognition: Participants had the opportunity to be recognized as Top Contributors within the OpenxAI community.
• Career Opportunities: The workshop served as a gateway for attendees to explore potential career paths with the OpenxAI team.
• This workshop marked a significant step in OpenxAI's mission to democratize AI development, providing developers with the tools and knowledge to create decentralized AI applications.

• Xnode Auth first version released
• NGINX auth_request module implementation supporting Ethereum messages and ip based access control
• Per domain path activatable and access list configurable through NixOS

Check how to integrate Xnode Auth in your app using the demo application

• XnodeOS will automatically issue {containerid}.container domains on the host DNS through DHCP leases
• These domains can be used for setting up reverse proxies to expose services running in a container
• Inter-container communication still uses multicast with .local domains

• Xnode Manager SDK library released for React

View on npm

• Xnode Manager SDK library released for TypeScript

View on npm

• Older changes are not being tracked on this changelog
• You can view older changes on GitHub.

• OpenxAI Mimic - Real-Time Voice Cloning
• OpenxAI unveiled Mimic at SuperIntelligence 2049 during Token2049 Dubai - a live demo that stunned the crowd with real-time voice cloning and witty, human-like AI responses, setting a new benchmark for human-AI interaction.
• The Technology Behind Mimic:
  • Bare Metal Servers → enabled split-second processing for real-time cloning.
  • XnodeOS → decentralized OS powering secure, scalable performance.
  • Ollama → orchestrated AI model deployment seamlessly.
  • Llama3 + Fetch.ai Agents → generated personalized, intelligent replies.
  • CSM-1b → text-to-speech engine capturing tone, pitch, and rhythm for lifelike voice.
• How It Worked (3-step process):
  • Speech Analysis → captured cadence, inflection, and tone.
  • Response Generation → created witty, natural replies via Llama3 + agents.
  • Voice Synthesis → Mimic delivered the reply instantly in the speaker’s own voice.
• Why It Mattered:
  • Showcased real-time human-AI interaction powered by decentralization.
  • Highlighted OpenxAI’s permissionless infrastructure (likened to “Skyscanner for GPUs”).
  • Proved AI can be both technically advanced and deeply engaging.
• Potential Use Cases:
  • Personalized AI assistants that sound like you.
  • Real-time multilingual voice translation.
  • Interactive education and training tools.
• OpenxAI’s Commitment:
  • Improving vocal inclusivity across diverse voice profiles.
  • Optimizing stability under varying network conditions.
  • Ensuring AI remains secure, scalable, and ethical.

On May 1, 2025, OpenxAI Network hosted SuperIntelligence 2049 at Dubai’s DAOs Hub during Token2049, co-organized with Fetch.ai, ASI Alliance, Aethir Cloud, SingularityNET, Daoshub, and 50xcap. The invite-only Crypto x AI summit showcased the transformative potential of decentralized AI, merging blockchain with autonomous intelligence to empower users, redistribute value, and challenge centralized platforms.

• Event Highlights:
  • Human Panel - “Decentralized Intelligence: Where Web3 Meets AGI”: Moderated by Gabriele Zennaro (50x Capital), the panel featured leaders including Humayun Sheikh (Fetch.ai), Mario Alberto Casiraghi (SingularityNET), Ashton Hettiarachi (OpenxAI), and Mark Rydon (Aethir). Discussions explored privacy, value redistribution, distributed AI economies, and community-driven innovation, emphasizing how OpenxAI’s infrastructure enables builders to capture value directly.
  • World’s First AI-Only Panel: Four autonomous AI personas - Tronald Pump, Bitter Schiff, Mickey Sailer, and Satoshi Mammoto - debated cryptocurrency and decentralization live, unscripted, powered by OpenxAI’s inference engine, XnodeOS, global bare metal servers, and Fetch.ai’s uAgents. The panel demonstrated fully autonomous, ideology-driven discourse, showcasing OpenxAI’s ability to deliver scalable, transparent, and engaging AI interactions.
  • Mimic Demo - Real-Time Voice AI: Attendees interacted with Mimic, a system capable of replicating voices in real-time using OpenxAI’s inference engine and Fetch.ai uAgents. The demo illustrated personalized, human-like AI interactions, highlighting decentralized, user-centric AI applications.
• Impact and Innovation:
  • OpenxAI’s technology enabled autonomous AI debate, ideological representation, and secure, scalable interactions.
  • The human panel underscored the redistribution of value in decentralized ecosystems, supporting privacy, sovereignty, and equitable AI-driven economies.
  • Partnerships amplified collaborative innovation, demonstrating the real-world applications of OpenxAI’s permissionless, decentralized infrastructure.

SuperIntelligence 2049 set a new benchmark for decentralized AI, highlighting OpenxAI’s leadership in building transparent, inclusive, and community-driven AI ecosystems. The event emphasized a future where AI empowers developers and users, reshaping industries through decentralized, autonomous, and ethical intelligence.

• OpenxAI Developer Workshop
  On April 5, 2025, OpenxAI hosted an exclusive developer workshop titled "OpenxAI Exclusive Developer Workshop" via Google Meet. The event, led by John Forfar from the OpenxAI Network, attracted 72 participants, including notable figures like Aman Chopra. The session aimed to introduce developers to OpenxAI's innovative approach to AI and Web3 integration.
• Workshop Highlights:
• Introduction to OpenxAI: The session commenced with an overview of OpenxAI's decentralized architecture, emphasizing its commitment to unrestricted access to AI models, decentralized compute through Xnodes, community governance via OpenxAIDAO, and the use of smart contracts for tokenization.
• Product Demonstration: A live demonstration showcased the OpenxAI Model Marketplace, decentralized compute network, governance tools, and tokenization features, illustrating how developers can leverage these tools to build and deploy AI models without centralized intermediaries.
• Interactive Q&A Session: The workshop concluded with a 30-minute Q&A session, allowing participants to engage directly with the host, ask questions, and explore how OpenxAI can enhance their AI development endeavors.
  This workshop underscored OpenxAI's dedication to fostering a community-led, permissionless AI ecosystem, empowering developers to innovate freely in the AI space.

• OpenxAI Developer Workshop
  On March 29, 2025, OpenxAI hosted an exclusive developer workshop in collaboration with Openmesh Network, focusing on integrating decentralized AI and Web3 technologies. Conducted via Google Meet, the session attracted 82 developers eager to explore the potential of a permissionless AI ecosystem.
• Workshop Highlights
• Introduction to OpenxAI: The workshop commenced with an overview of OpenxAI's mission to democratize AI through unrestricted access to models, decentralized compute via Xnodes, community governance via OpenxAIDAO, and smart contract-based tokenization.
• Product Demonstration: A live demonstration showcased the OpenxAI Model Marketplace, decentralized compute network, community governance features, and tokenization mechanisms, illustrating how developers can leverage these tools to build and deploy AI models without centralized intermediaries.
• Interactive Q&A Session: The workshop concluded with a 30-minute Q&A session, allowing participants to engage directly with the host, ask questions, and explore how OpenxAI can enhance their AI development endeavors. This workshop underscored OpenxAI's commitment to fostering a community-led, permissionless AI ecosystem, empowering developers to innovate freely in the AI space.

• Progress Update
• Engineering Updates
  • Enhanced Efficiency: Achieved 96% efficiency in AI model deployment on bare metal.
  • Cost Optimization: Reduced compute costs by 5% through Xnode refinements.
  • Global Infrastructure: Scaled to 350 G/F GPUs, 28 PB memory, 950 PB bandwidth, across 500 cities and 35 providers.
  • Developer Access: Expanded the bare metal demo pool with new testing features.
  • Compute Cost Transparency: Maintained $0.12/compute hour and $0.02/storage TB with forecasting tools.
  • Performance Monitoring: Upgraded real-time analytics with predictive modeling for efficiency.
• New AI Models Added
  • SmolLM (HuggingFace): Integrated with 135M, 360M, and 1.7B parameters, excelling in reasoning and knowledge tasks.
  • Qwen: Added versatile variants including Qwen-Max and Qwen-Plus for text generation and programming.
  • Gemma: Incorporated lightweight model for text generation and fine-tuning, boosting developer productivity.
• New AI Agents Added (Beta)
  • Fetch.ai uAgents: Added lightweight framework for building decentralized AI agents that communicate and transact.
  • Each agent had a unique identity from its seed phrase, supported local/remote communication, and performed scheduled tasks.
  • The feature was released in beta, with access available via developer channels.
• Community Event
  • OpenxAI Workshop (March 21, 2025): Hosted session on creating and deploying AI models with OpenxAI Studio.
  • Covered practical steps for building and monetizing models like DeepSeek and ChatGPT.
  • Received overwhelmingly positive feedback, with attendees highlighting improved understanding of decentralized AI.
• Launch of Genesis
  • OpenxAI Studio: Enhanced with new UI features for model tuning and real-time collaboration.
  • Model Integration: Added 5 new open-source AI models, fueling decentralized innovation.
  • Xnode & Bare-Metal: Optimized compute allocation, enabling 20% more concurrent workloads.
  • OpenxAI Dashboard: Released advanced analytics to track compute usage, project milestones, and funding progress.
• Community Updates
  • Growth Milestone: Our open community surpassed 12K active members.
  • Zealy Campaign: We added new tasks for model deployment and bug reporting.
  • 100x Community Content Campaign: Concluded on March 24, 2025, with nearly 600 creators submitting content and $50K in rewards distributed.
  • 30 Devs in 30 Days: We invited talented developers to join our decentralized team and shape the future of open AI.

• OpenxAI × MCP: LLMs Talking Ethereum, No Middlemen
• OpenxAI is advancing the decentralization of AI by integrating MCP (Machine Client Protocol) with Ethereum smart contracts and its Xnode decentralized compute network. MCP, an open-source protocol introduced by Anthropic, standardizes how AI models interact with external systems - acting like a universal connector that enables AI to access data, execute functions, or perform transactions without custom integrations.
• By hooking MCP into Ethereum, OpenxAI models can interact directly with smart contracts in a trustless, permissionless manner. Models can mint NFTs, manage fractionalized datasets, process payments in OPENX tokens, and execute complex logic - all without centralized intermediaries. This architecture leverages blockchain standards including:
  • ERC-721 / ERC-6551: Tokenizing AI models as unique, upgradeable NFTs with smart contract wallets.
  • ERC-1155: Fractionalized ownership of datasets and training sets.
  • ERC-20: Fungible tokens for payments and staking-based AI service access.
  • ERC-4337: Gasless transactions and subscription payments.
  • EIP-2981: NFT royalties, ensuring ongoing compensation for AI developers.
  • Decentralized Storage: IPFS, Arweave, or Filecoin ensures censorship-resistant, tamper-proof AI models.
• Key Use Cases:
  • Your AI, Your Rules: Mint AI models as NFTs on Xchange, control licensing via MCP clients, and maintain global ownership without centralized gatekeepers.
  • Crowd-Powered Data Riches: Fractionalized datasets (ERC-1155) allow contributors to monetize data directly while AI models tap it via MCP for live insights.
  • Talk to AI, Pay with Ease: Voice-to-voice AI apps on XScape charge OPENX tokens or subscriptions gas-free through ERC-4337.
  • Royalties That Keep Giving: Developers earn continuous royalties on NFT trades and licensing via EIP-2981.
  • Smart Trading, No Masters: XForge trading bots operate autonomously, using real-time Ethereum data and decentralized compute, fully owned by the creator.
• Implementation Highlights:
  • Xnodes: Lightweight NIX-based unikernel OS for decentralized compute.
  • Smart Contracts: ERC-721/ERC-6551 for AI NFTs, ERC-1155 for dataset shares, ERC-20 for payments, and ERC-4337 for gasless subscriptions.
  • MCP Servers & Clients: AI clients query MCP servers running on Xnodes to interact with Ethereum contracts seamlessly.
  • Model Storage & Metadata: AI models stored on Xnodes, with NFT metadata pinned for integrity and provenance.
  • Playground Apps: Interactive apps leverage MCP clients to mint models, fetch data, and execute AI logic in real-time.
• The Bigger Picture:
  • This initiative reasserts the original ethos of Web3: AI and blockchain working together without middlemen. OpenxAI’s integration of MCP, Ethereum standards, and Xnodes enables fully decentralized AI ownership, access, and monetization, giving developers control over their creations and income streams. By removing reliance on centralized cloud providers, OpenxAI fosters a fair, open, and resilient AI ecosystem, empowering creators and communities alike.

• The Decentralized Version of Open Hugging Face is Now Live
• The AI Ecosystem Giant: Hugging Face
  • Founded in 2016, Hugging Face grew into the largest AI hub, hosting 200K+ models.
  • Serves 2M+ developers with tools for model hosting, fine-tuning, and deployment.
  • Valued at $2B+, often called the “GitHub of AI.”
  • Centralized structure made AI more accessible but also introduced barriers.
• Why Hugging Face Succeeded
  • Solved AI fragmentation by centralizing models in one place.
  • Open-source ethos accelerated collaboration and breakthroughs.
  • Tools like Transformers library became industry standards.
• The Cracks in Centralization
  • Permissioned access limits participation in restricted regions.
  • High costs for advanced features price out small teams.
  • Privacy risks from centralized data hosting.
  • No direct rewards for model creators.
• Enter OpenxAI: The Decentralized Challenger
  • Permissionless & community-owned, powered by Xnodes and smart contracts.
  • No accounts, credit cards, or corporate gatekeepers required.
  • DAO governance ensures decisions come from the community.
  • Models are tokenized assets, letting creators monetize directly.
• Hugging Face vs. OpenxAI: Key Divergences
  • Access: Open, borderless vs. account-gated.
  • Governance: DAO vs. executives.
  • Compute: Decentralized Xnodes vs. corporate clouds.
  • Monetization: Tokenized rewards vs. none.
  • Censorship: Resistant vs. centralized control.
• Why OpenxAI Matters
  • Cuts compute costs by up to 80%.
  • Levels the playing field for global developers.
  • Keeps data private, censorship-proof, and community-owned.
  • Positions AI as a commons, not a corporate monopoly.