At Intelligence Cubed (i³), our innovation is guided by an elite group of fellows from premier institutions. This powerful blend of academic rigor and industry savvy pushes the boundaries of what’s possible in the Modelverse — where AI models become on-chain IP with programmable licensing and lineage royalties.

This esteemed group includes Xuandong Zhao (Postdoc, UC Berkeley RDI, supervised by Prof. Dawn Song), Yuejiang Liu (Postdoc, Stanford AI Lab), Yaqi Xie (Postdoc, CMU Robotics Institute), Jason Dou (Postdoc, Harvard), and Peter Wang (Postdoc, Caltech), Shiyi Du (Ph.D., CMU, Computational Biology), Jiayuan Liu (Ph.D., CMU, Computer Science), Yitong Li (Ph.D., Stanford, Computational Science), and Chengfeng Mao (Ph.D., MIT, Computer Science).

Here are the introductions to our Fellows:

Xuandong Zhao

• Postdoctoral researcher at UC Berkeley with BAIR and Berkeley RDI, working with Prof. Dawn Song
• PhD in Computer Science at UC Santa Barbara, advised by Yu-Xiang Wang and Lei Li; BS in Computer Science at Zhejiang University
• Research in machine learning, natural language processing, and AI safety, focusing on responsible and reliable generative AI
• Published at leading venues including ICML, ICLR, NeurIPS, ACL, and IEEE S&P — with recent work at ICML 2025, ICLR 2025, IEEE S&P 2025, NeurIPS 2024 • Highlights include SoK: Watermarking for AI-Generated Content (IEEE S&P 2025), An Undetectable Watermark for Generative Image Models (ICLR 2025), and Invisible Image Watermarks Are Provably Removable (NeurIPS 2024)
• Industry experience at Alibaba, Microsoft, and Google
• Recognitions include Rising Star in AI (2025) and Rising Star in Adversarial ML (2024), plus multiple fellowships and scholarships

Personal page: https://xuandongzhao.github.io
Google Scholar: https://scholar.google.com/citations?user=CxeH4uoAAAAJ&hl

Yuejiang Liu

• Postdoctoral fellow at Stanford working with Prof. Chelsea Finn
• PhD in Computer Science at EPFL advised by Prof. Alexandre Alahi; research internships with Francesco Locatello, Chris Russell, and Bernhard Schölkopf
• Research in brain-inspired/causal representation learning, data curation, and test-time adaptation for embodied robots
• Published at NeurIPS, ICCV, ICML, ICLR, CVPR, CoRL, and RSS
• Honors include SNSF Postdoc Fellowship (2024), National Doctorate Award (2022), and Best Paper at RSS RoboReps (2025); competition wins ICML BoltzGo 1st (2021) and ICCV TrajNet++ 1st (2021)

Personal page: https://sites.google.com/view/yuejiangliu/home
Google Scholar: https://scholar.google.com/citations?user=Xi-B5WIAAAAJ&hl=en

Yaqi Xie

• Postdoctoral scholar at Carnegie Mellon University in the Robotics Institute, advised by Prof. Katia Sycara.
• PhD in Computer Science at the National University of Singapore under Prof. Harold Soh.
• Research focus in Human-AI Synergy through neural-symbolic fusion, developing interpretable, robust, and trustworthy AI for perception, decision-making, and generative models.
• Published at leading venues including NeurIPS, CVPR, ICRA, IJRR, and HRI with recent work accepted at NeurIPS 2024, CVPR 2024, ICLR 2025, ICCV 2025, ACL 2025, AISTATS 2025, and WACV 2025.
• Highlights include “Embedding Symbolic Knowledge into Deep Networks” (NeurIPS 2019), “Dual Prototype Evolving for Test-Time Generalization of VLMs” (NeurIPS 2024), HiKER-SGG (CVPR 2024), and ONLY (ICCV 2025).
• Recognitions include Best Paper at RSS GenAI-HRI Workshop 2025 and Best Paper at an ICRA Workshop in 2024.

Personal page: https://yaqi-xie.me
Google Scholar: https://scholar.google.com/citations?user=lBCCo0EAAAAJ&hl=en

Yitong Li

• PhD at Stanford University, specializing in computational science, inverse problems, and machine learning for Earth systems
• Co-teaches Stanford courses: Computations in Civil and Environmental Engineering and Imaging with Incomplete Information
• Research focus in interpretable, physics-informed models for subsurface imaging and environmental sensing
• Co-authored a multi-institutional study aligning large language models with the UN’s Sustainable Development Goals
• Published and cited in Earth system modeling and AI, aiming to make environmental modeling more transparent and impactful

Google Scholar: http://scholar.google.com/citations?user=SufajHwAAAAJ

Chengfeng Mao

• PhD student at MIT Sloan in Marketing Science; prior MS in Computer Science from CMU and BS in Computer Engineering from UIUC
• Professional experience includes building large-scale data pipelines at Yahoo, Cask (CDAP), and Google Cloud, shaping his pursuit of scalable causal analytics
• Research bridges machine learning, NLP, and marketing science from multimodal transformers to concept mining and LLM-based customer-needs extraction for managerial insight.
• Published at ACL on Integrating Multimodal Information in Large Pretrained Transformers (MAG) and released open-source code; work spans multimodal sentiment and adaptation.
• Recent work in Information Systems Research and Marketing Science explores how AI can extract managerially relevant concepts and predict consumer needs

MIT Sloan profile: https://mitsloan.mit.edu/programs/phd/chengfeng-mao Google Scholar: https://scholar.google.com/citations?user=NfilsZkAAAAJ&hl=en

Jason Xiaotian Dou

• Postdoctoral Research Fellow at Harvard University, focusing on AI for healthcare and biology. Currently also the Founder of Marbella AI, incubated at Harvard Innovation Labs, bringing generative AI solutions to traditional industries.
• PhD in Computer Engineering from the University of Pittsburgh, applying machine learning to healthcare, biology, finance, mobile, social science, and operations.
• Master’s in Information Science from Cornell University and B.S. in Computer Science from Peking University.
• Research interests span representation learning, computational oncology, causal machine learning, multimodal learning, reinforcement learning, AIoT, and knowledge graphs.
• Published in AAAI, NeurIPS, ICML, ML4H, ICIBM, Clinical Cancer Research, with work featured in major outlets including Tencent, Sina, and Caixin.

Personal page:https://sites.google.com/site/douxiaotianjason/home?authuser=0
Harvard profile: https://sites.harvard.edu/xiaotian-dou/#
Google Scholar: https://scholar.google.com/citations?hl=en&user=c-VrJvIAAAAJ&view_op=list_works&sortby=pubdate

Jiayun (Peter) Wang

• Postdoctoral Researcher in Computing & Mathematical Sciences at Caltech, advised by Prof. Anima Anandkumar, working on AI for science and healthcare with applications in lung and brain imaging (ultrasound, photoacoustic tomography).
• PhD in Vision Science (Computer Vision Track) from UC Berkeley, advised by Prof. Stella X. Yu and Prof. Meng C. Lin. Research focused on structure-aware representation learning and its applications to healthcare.
• B.S. in Electrical Engineering from Xi’an Jiaotong University, recognized with the National Scholarship of China and Outstanding Graduate Award; spent 2017–2018 as a visiting student at UC Berkeley.
• Research expertise spans representation learning, inverse problems, 3D vision, and AI for healthcare. Contributions include ECCV 2024 Oral Presentation, Best Paper at ML4H (2023 & 2024), CVPR PBVS Best Paper (2019), and publications at TPAMI, CVPR, ICCV, ECCV.
• Honors include the Vector Institute Fellowship (2023 offer), HKSTP Best Paper Award (2019), and multiple top international recognitions.

Personal page: https://pwang.pw
Google Scholar: https://scholar.google.com/citations?user=IBn7PdYAAAAJ

As the Model Context Protocol (MCP) gains unprecedented momentum, with over a thousand MCP-compliant models now live on platforms such as Smithery — we are witnessing a foundational shift in artificial intelligence development. This new paradigm is not just about modularity or interoperability; it is fundamentally about redistributing control over AI development and access.

At Intelligence Cubed (i³), we are not merely participants in this evolution, we are its infrastructure layer. We are building the tools, incentives, and governance mechanisms required to support a future in which model ownership, attribution, and discovery are equitable, open, and democratic by design.

We are developing a decentralized, tokenized modelverse tailored for seamless MCP integration. In this ecosystem, any developer can deploy, monetize, and govern their MCP-compatible models without intermediaries or restrictive gatekeepers. Our approach centers on fair attribution, transparent usage metrics, and programmable economic rewards, all under the watch of a community-governed DAO.

The Expanding Role and Promise of MCP

The Model Context Protocol is more than a standard; it is an architectural reimagining of how AI models communicate and scale. By formalizing contextual metadata, MCP enables:

• 🔄 Interoperability between disparate agents and systems
• 🧱 Composable architectures for modular, stackable intelligence
• 🚀 Faster integration across developer ecosystems

As outlined in emerging technical standards, including from Metaphor Systems, MCP offers a universal schema for model description and contextualization — laying the groundwork for federated model ecosystems with no central broker.

This has profound implications: AI development no longer needs to reside in isolated silos or under proprietary stacks. It can become a permissionless, global dialogue, fueled by common language and interoperable logic.

How Intelligence Cubed Operationalizes MCP in a Decentralized Context

At Intelligence Cubed, MCP is not a future consideration, it is a live standard. We’ve built our infrastructure natively around it, meaning:

• ✅ All MCP-compatible models are deployable without centralized approval
• 💡 Model creators can tokenize usage rights and governance shares
• 📈 Models are plugged into a DAO-supervised benchmark and curation layer
• 🧠 No blockchain experience is required for participation — our UI abstracts it away

From a single researcher uploading their first model to a research lab managing 50+ deployments, i³ ensures models are treated as economic and creative assets, not just utilities.

Why Now? The Critical Window for Decentralized AI

The timing for decentralized infrastructure could not be more pressing:

• 📊 Smithery’s repository now exceeds 1,000+ MCP models
• 📈 Developer tooling demand is rising for composable, context-aware agents
• 💸 Users are increasingly disillusioned with flat-rate SaaS pricing and are seeking pay-as-you-go alternatives

Yet key bottlenecks persist:

• No native monetization for model builders, many models accrue visibility, but not value.
• Opaque discovery and ranking systems dominated by closed-source logic.
• Rigid pricing models where users pay for bundled features they don’t use.

Intelligence Cubed resolves these bottlenecks by embedding decentralized economics directly into the modelverse.

We provide:

• 🔗 On-chain attribution primitives for fair compensation
• 🔍 Transparent usage data to benchmark quality and utility
• 📡 A DAO-run governance mechanism for upgrades, rewards, and discoverability

This is not just decentralization for its own sake. It is decentralization in service of economic participation, pluralistic evaluation, and model diversity.

Why Intelligence Cubed Is a Unique Fit for This New Era

• 🧬 We abstract blockchain complexity for AI developers, no wallet management or contract coding required
• 🛠️ Our stack is built for modularity, not monolithic model hosting
• 🎯 We support every phase of the model lifecycle, from deployment to monetization, benchmarking to remixing
• 🧠 Our DAO is not symbolic, it governs model ranking, royalties, and open-sourcing milestones

We transform MCP from an integration point into a living protocol for ownership, economics, and coordination.

A Call to Builders, Thinkers, and Innovators

The next chapter of AI will not be defined by a single company or standard. It will be built collectively, by communities who believe in modularity, openness, and economic fairness.

If you believe that the future of AI should be designed for the people, not monopolized by platforms — then Intelligence Cubed is the infrastructure you’ve been waiting for.

🔗 Explore: https://icubed6.godaddysites.com

📃 I Cubed Litepaper: https://intelligence-cubed.gitbook.io/intelligence-cubed/i-cubed-litepaper

📩 DM us to list your MCP model or request early access

📢 Follow: https://x.com/I3_Cubed

Let’s build a new AI commons — one model, one creator, one vote at a time.

About I Cubed

I Cubed pioneers decentralized intelligence empowerment through AI model ownership democratization. At the forefront of redefining AI creation and distribution, we empower next-generation creators to maximize model exposure while ensuring fair valuation and sustainable monetization of intellectual contributions.

Vision

To establish a decentralized AI ecosystem that drives innovation, ensures fairness, and enables shared prosperity through tokenized intelligence assets.

We envision a future where cutting-edge AI models become universally accessible while enabling every creator to sustainably profit from their innovations.

Team

I Cubed’s technical team features top talents from Berkeley, Stanford, CMU, and Duke. The founding team authored over 30 publications in top AI journals and conferences like IEEE and ICLR. Through our innovative platform, cutting-edge language models are made accessible to researchers, developers, and enthusiasts worldwide.

Preface

The AI market suffers from extreme centralization, with companies like OpenAI and others capturing a significant share of users. For instance, OpenAI’s ChatGPT alone is estimated to hold approximately 60% of the conversational AI user base as of early 2025, with over 400 million weekly active users. When people think of AI, their instinct is to turn to ChatGPT or similar offerings from tech behemoths, overshadowing specialized models painstakingly developed by small teams or independent researchers for niche applications. As a founding team — having published research on deep neural architectures and built machine learning models ourselves — we have experienced this frustration firsthand: the market’s bias toward established players often leaves innovative, tailored solutions unnoticed.

Training a competitive language model demands immense resources, creating a formidable barrier to entry. For example, OpenAI’s GPT-4o, an advanced multimodal model, is rumored to have cost upwards of $100 million to train, factoring in computational resources, data acquisition, and engineering efforts. Similarly, DALL-E 2, another Transformer-based model from OpenAI introduced by Ramesh et al. (2021), boasts 12 billion parameters and was trained on over 400 million captioned images, requiring significant computational power. While OpenAI bore the costs of training DALL-E, they controversially decided against open-sourcing the model, meaning the code and architecture are not publicly available. Smaller models remain inaccessible to most, A 7B parameter LLM requires about 100,000 GPU hours ($150,000 on Nvidia A100). These high resource demands disproportionately favor well-funded organizations, sidelining independent innovators.

However, a shift began in January 2025 with the emergence of DeepSeek, a Chinese company that disrupted the AI landscape. DeepSeek unveiled their MoE model (DeepSeek V3) achieving GPT-4 level reasoning at approximately 5.5% training cost. Their FlashMLA improves the performance of NVIDIA H800 AI chips by 8 times. Inspired by this breakthrough, an explosion of efficient training techniques and products follows:

• Stanford’s LIRE framework enables 26-minute LLM training via dynamic architecture search.
• A proliferation of derivatives and adaptations fostered by DeepSeek’s open-sourcing of its models, with communities rapidly building on its frameworks.
• The rise of edge AI products, spurred by DeepSeek’s efficient models that can be deployed on consumer hardware, enabling sophisticated AI to run locally without cloud dependency (alliedinsight.com, February 14, 2025).

The U.S. AI market is shocked. These breakthroughs signal the dawn of low-cost AI development and suggest an impending explosion in affordable AI models, driven by open-source tools and cost-effective training methods.

• Yet 68% end-users still pay $20+/month for generic subscriptions (AIProductBench)

Despite DeepSeek’s promise for developers, the consumer market lags behind. A predominance of AI users still relies on subscription-based large models, spending at least $20 monthly, with an estimation of 20–30% subscribing to multiple services.

Meanwhile, 40% of surveyed businesses report a lack of AI solutions tailored to their needs, highlighting a disconnect between supply and demand. No mature platform yet exists to bridge this gap, allowing users to “vote with their feet” and connect developers with those who need their models. While open-source platforms like Hugging Face foster collaboration and learning, they fail to provide creators with direct revenue. Our interviews with PhD students from leading universities revealed a common reluctance to open-source models they’ve invested significant time and computational resources into, due to the absence of financial incentives.

OpenAI’s 2019 pivot to a for-profit model, followed by 2023’s leadership crisis exposing internal clashes between commercialization and safety, starkly deviated from its original nonprofit mission of open, ethical AI for all. This corporate turbulence underscores the risks of centralized control over AI development, contrasting with our web3 product’s vision: decentralized, open-source models that democratize access and uphold transparency, aligning with OpenAI’s founding ideals.

Decentralized blockchain technology offers a potent solution to these challenges. Unlike traditional AI models controlled and trained by giants, deploying models on a decentralized network enables developers and users to collaboratively train, improve, and govern them. This reduces reliance on centralized entities, lowers barriers to entry, and creates a transparent, equitable ecosystem where value is shared among contributors. By tokenizing AI models and leveraging community-driven mechanisms, I Cubed aims to unlock this potential, redefining how AI innovation is nurtured and rewarded.

Problems

The creation of a marketplace for AI models introduces a mix of traditional marketplace challenges and unique obstacles tied to the nature of AI assets.

Marketplace Liquidity Challenges

A marketplace for AI models encounters the same supply-and-demand hurdles as any emerging platform, amplified by the specialized characteristics of AI. A primary challenge is the cold-start problem: for the marketplace to grow, there must be sufficient liquidity — , a balanced supply of models and demand from users — from the very beginning. To attract AI model creators and stimulate participation, the platform must offer clear, usage-based rewards. This requires tracking model usage accurately and distributing proportional payments to providers, incentivizing them to contribute high-quality models.

Trustless Verification Complexities

To establish a trustless network with robust economic incentives, the marketplace must ensure that models are genuinely used and deliver high quality. In the absence of centralized oversight, verifying model performance and usage becomes a significant challenge. Without an effective verification system:

• Creators could submit low-quality or untested models, undermining the platform’s credibility.
• Users might falsely report usage to manipulate rewards, disrupting the economic model.
• The lack of quality control could flood the marketplace with substandard contributions, reducing its overall value.

A reliable mechanism to measure usage (e.g., through tracked interactions like API calls or downloads) and assess quality (e.g., via performance metrics or user feedback) is critical to maintaining trust and ensuring rewards reflect real contributions.

Neutral Evaluation

Exceptional AI models deserve sufficient exposure — especially to their intended audience — and fair assessment of their value. In many platforms, visibility is dictated by algorithms or editorial decisions, which can introduce bias and favor well-known contributors. To address this, a neutral voting system is necessary, enabling impartial evaluation of models based on merit.

Without such a system:

• High-quality models from emerging creators may remain unnoticed, reinforcing the dominance of established players.
• Unfair or opaque evaluations could distort perceptions of model quality, discouraging participation.
• Lack of transparency erodes trust, a vital component of any marketplace.

A community-driven, neutral evaluation process ensures that models are judged fairly, promoting diversity and innovation across the platform.

Solution

Overview

I cubed is building a decentralized, community-driven marketplace for AI model development and usage, powered by blockchain technology for transparency and democracy. It offers AI model creators a platform to earn income and reputation through their creation and partial ownership transfer, while the users can get rid of prevalent expensive subscriptions from large companies, search for the best models in niche areas, only pay for their usage, and stake the models they look to promising future. Governance is managed by the DAO to ensure trust and fairness for all participants.

For Model Creators

Unlike Web2 platforms where creators work under centralized ownership without recognition or sustainable income, I Cubed provides a Web3-native infrastructure where creators retain ownership and tokenize their AI models as digital assets.

Developers can:

• Deploy fully on-chain autonomous AI models validated via Proof of Intelligence.
• Monetize through create-to-earn rather than “share-for-free”.
• Retain partial model ownership while transferring shares to fund development.
• Benefit from exposure through a community-curated recommendation engine, promoting model discovery.
• Compete fairly in an open, DAO-driven AI benchmark that highlights performance through community voting.

For Community Users

Community users can:

• Discover and use high-quality, niche models without committing to large, flat-fee subscriptions.
• Stake tokens in models they believe in — earning returns if the models perform well.
• Remix, fork, and experiment with open-source models, contributing to an evolving ecosystem.
• Participate in democratic governance by voting on benchmarks, model evaluations, and development proposals.
• Become active co-creators rather than passive consumers.

Layers of I Cubed

Key Workflow Example

Decentralized AI modelverse

I Cubed adopts a Hugging Face–like model space framework while fundamentally rethinking its architecture through decentralization and tokenized ownership. Traditional Web2 platforms often rely on creators voluntarily open-sourcing their work to gain visibility or handing over ownership for compensation. In contrast, I Cubed ensures that creators retain intellectual property rights and earn income through usage-based rewards — creating long-term incentives for sustainable innovation and community self-governance.

This approach transforms the “open-source for exposure” model into a “create-to-earn” ecosystem. As models perform well, they naturally attract more usage and community-driven staking to unlock open-source access. Developers not only gain visibility and reputation but also earn from each usage, unlike on platforms like Hugging Face where models are shared without direct monetization.

At the same time, consumers benefit from pay-per-use access to niche, high-quality models — without committing to expensive multimodal subscriptions from tech giants or overpaying for one-time model access.

Key Differences from Hugging Face

Proof of Intelligence

Proof of Intelligence (PoI) is I Cubed’s decentralized verification mechanism designed to ensure AI models are both original and performant while remaining trust-minimized and market-driven. Instead of executing models on-chain, PoI leverages off-chain compute environments — including Trusted Execution Environments (TEEs) and decentralized inference nodes — to securely evaluate model behavior.

Model originality is verified through functional deduplication, while performance is attested using benchmark summaries, optionally wrapped in zero-knowledge proofs for privacy. Usage metrics are aggregated by oracles to trigger rewards and dynamically adjust pricing. Community members can report fraud or plagiarism, with slashing enforced through a lightweight challenge system.

By treating intelligence as a verifiable resource — validated through TEEs, oracles, and ZK attestations — PoI enables a scalable, trust-minimized framework where creators earn based on quality, not visibility alone.

Initial Model offering

I Cubed introduces a novel Initial Model Offering (IMO) mechanism that enables AI models to be offered, traded, and collectively owned — much like early-stage equity in a startup. This process not only helps creators monetize their work early but also aligns community incentives around promising AI systems.

Process

Model Submission

The model creator uploads a model to the I Cubed modelverse and initiates an IMO request.

Valuation

The creator sets an initial token price based on expected demand.

I Cubed’s recommendation engine offers guidance using traffic data and similar model benchmarks.

IMO Staking

Users stake into the IMO pool to reserve ownership shares at a fixed price.

Shares can be traded among participants, but cannot be withdrawn or refunded (anti-rollback rule).

Open-Sourcing Trigger

Once 51% ownership is staked, the model enters a 1-day soft lock until 00:00 PT (T+1 business day).

On the second business day, the model is open-sourced on IPFS, and pricing is released to market.

Secondary Market Trading

Stakeholders can now trade model tokens freely. New participants can enter, trade, or accumulate ownership based on market dynamics, similar to post-IPO equity trading.

Model Valuation Recommendation Engine

Each model entering the IMO pool requires an initial valuation set by the creator. To guide this decision, I Cubed offers a recommended price range using a dynamic valuation algorithm inspired by secondhand marketplaces and recommender systems.

Where:

Anti-Rollback & Lock-in Rules

To maintain pricing integrity and prevent speculative gaming, the IMO process enforces a non-refundable participation policy:

• During the IMO window, users can freely trade their pre-staked shares (at fixed price) with others.
• However, redemptions or refunds to the IMO contract are strictly prohibited — ensuring capital commitment is real.

Post-IMO Market Dynamics: Equity-like Ownership & Liquidity

Once the 51% ownership threshold is reached, the stake pool is locked until 00:00 PT on the same business day. At 00:00 PT on the next business day, the model is:

• Open-sourced via IPFS and verified by the platform
• Unlocked for full trading, enabling secondary ownership transfers

Stakeholders can now:

• Trade their shares freely with others on the marketplace
• Offer partial sales (like equity vesting schedules or unlisted option grants)
• Benefit from price appreciation as usage, exposure, and community trust grow

This mimics a pre-IPO equity structure where stake tokens function as non-public equity pre-open-source. Once open-sourced, the tokens become liquid and act like public stock.

Neutral Evaluation & Democratic Pricing

Evaluating AI models across diverse domains is inherently difficult — especially when comparing niche single-purpose models with large, multimodal systems. Multimodal models often gain outsized popularity due to their broad utility, overshadowing specialized models that perform exceptionally well in focused tasks but lack mass visibility. To counter this imbalance, the I Cubed modelverse introduces a diversified recognition and evaluation framework.

Instead of simply asking, “How many people use Model X?”, we propose a more meaningful question:

“How many critical outcomes uniquely depend on Model X?” This reframing highlights functional importance over superficial popularity.

In a well-functioning system, there should be a mechanism where deriving a result is complex, but verifying it is simple. Futarchy, a voting mechanism that was originally introduced by Robin Hanson as “vote values, but bet beliefs”, fits naturally in this decentralized system.

(Briefly explained: this mechanism selects a set of goals — which can be any measurable metrics — and combines them into a target metric M. When a decision needs to be made (say, a YES/NO outcome), three prediction markets are opened:

(i) whether YES or NO will be chosen;

(ii) the value of M if YES is chosen (otherwise 0);

(iii) the value of M if NO is chosen (otherwise 0).

From these, the system can infer which decision the market believes will be more beneficial for M.)

Thus, we bring all community participants into this system to vote values and bet beliefs. I Cubed’s crowdfunding mechanism aims to let the network transparently and fairly evaluate your impact. Once a model enters the Initial Model Offering (IMO) pool, 51% of ownership is opened for public staking. Anyone from the community can stake on models they personally believe to be promising. When 51% has been fully staked, ownership is temporarily locked, the model is open-sourced, gains broader exposure, and stakeholders can further trade their stakes and earn dividends.

The more people rely on and recognize your model, the more likely it is to be open-sourced through community crowdfunding — and its valuation will rise with market expectations. This approach ties model evaluation to both market activity and democratic consensus, turning pricing power into a collective decision.

How does this mechanism reward and encourage “correct” behavior? Those who stake early on models that later prove impactful will gain exposure, dividends, and token appreciation. Those who stake on poor-performing or low-utility models will see minimal returns or incur opportunity costs. This mechanism creates built-in incentives for truthful signaling and discourages speculation or manipulation.

Our democratic benchmark system ensures that well-performing vertical models receive proper exposure. Every niche model can be surfaced and discovered — not just high-profile general-purpose models. This prevents vertical excellence from being neglected in favor of popularity alone.

Decentralization

Why Web3 Is the Only Viable Path for Create-to-Earn

If we truly want creators and developers — from any domain — to create and earn through AI model sharing and innovation, Web3 is not just an option — it’s a necessity. The Web2 framework is fundamentally flawed for this use case. Earnings are often locked within walled platforms, where withdrawal is restricted, slow, or taxed by excessive fees. For most creators, liquidity is low, value is non-transferable, and monetization is at the mercy of gatekeepers.

We see this failure across creative sectors. In both gaming and AI, developers dedicate time and expertise to build impactful products, only to receive non-cashable points or exposure that never translates into real income. Without reliable monetization, even great projects can’t survive — and neither can their communities.

Web3 introduces a fundamentally different system. Through tokenized ownership, permissionless earning, and portable value, creators are empowered to earn with clarity, autonomy, and aligned incentives. It’s not just about making money — it’s about restructuring economic logic at the protocol layer, enabling a creator-first innovation economy to emerge.

Decentralization as an Economic and Moral Imperative

Today’s AI ecosystem increasingly resembles early industrial monopolies: power is concentrated, access is gated, and platforms extract value while offering little in return. A handful of tech giants determine who builds, who profits, and who gets seen. This centralization not only limits innovation but replicates the inequities of the Web2 era — where users and builders alike have minimal agency.

Web3 reimagines this structure. Just as 20th-century companies evolved from robber-baron monopolies into employee-owned firms with equity incentives, Web3 ushers in a new era of distributed ownership and governance. In this model, participants are no longer passive users — they are stakeholders. I Cubed embodies this vision through model tokenization, pay-per-use mechanics, and DAO-based governance, where value and voice are shared, accountable, and transparently coordinated.

Rebuilding the Order of the AI Industry

The AI industry’s existing power dynamic is broken and unsustainable. It slows innovation, restricts access, and rewards only the few. Conversations around AI safety and alignment are important — but so are the economic rights of small creators, the decentralization of training data, and infrastructure access for independent contributors.

Today’s model mirrors a digital feudal system, where big tech and VC-backed startups act as lords, and everyone else builds under their terms. As Vitalik Buterin noted, AI must evolve into a space where every actor — developers, researchers, users — can be self-sufficient, democratic, and organically networked into a regenerative, open community.

This shift isn’t just moral — it’s structural. For AI to scale as a sustainable industry, we must rebuild its economic order — redefining how value is created, distributed, and governed.

The Revival of Distributed Capital and Intelligence

Decentralization is not merely a tech shift — it’s the return of capital democracy. Just as public equity markets allowed millions to co-own and benefit from corporations, I Cubed enables communities to co-own intelligence itself. A model is no longer a static asset locked in a corporate vault — it. It becomes a living, evolving protocol, backed by belief, utility, and market validation.

Ownership fragmentation isn’t a weakness — it’s a strength. When thousands of people stake in a model’s success, the system gains momentum, accountability, and resilience. Open-source models become public goods powered by private incentives.

I Cubed revives the best parts of Web2 — collaboration, iteration, scale — while rejecting its worst: gatekeeping, exploitation, and central control. True AI democratization doesn’t begin with better models. It begins with fairer systems to own, govern, and grow them — together.

Future Ecosystem

To enable a truly decentralized and sustainable AI model economy, I Cubed is building an end-to-end infrastructure that supports not only model ownership and usage, but also scalable deployment, cost-efficient compute, and privacy-preserving execution. Our future ecosystem consists of three key components:

Hardware Membership

In the near future, I Cubed will introduce proprietary hardware equipped with Trusted Execution Environments (TEE). These edge devices serve as gateways to a membership-based system, allowing users to run AI models they own or license directly on secure hardware. By executing models locally, these devices reduce dependency on centralized hosting, lower inference costs, and enhance privacy by keeping user data within the device. If misuse or tampering is detected, devices can be remotely suspended, enforcing access control without compromising decentralization. This model strengthens data sovereignty while offering creators a controlled and protected runtime environment — paving the way for distributed AI compute infrastructure at the edge.

Decentralized Compute Power Integration

Upstream, I Cubed will collaborate with decentralized GPU platforms to create a shared computing layer that dynamically allocates resources based on real-time model demand. Through partnerships with networks such as Akash or io.net, the platform will enable cost-efficient compute for model training, fine-tuning, and on-demand demos without relying on traditional cloud monopolies. This flexible compute structure reduces overhead for creators and ensures the modelverse remains operationally scalable and permissionless.

Inference Endpoints & Public Demo Spaces

At the application layer, I Cubed will provide seamless tools for deploying AI models into production. Developers can convert any model in the modelverse into live inference endpoints, host public demos, or integrate models into real-world workflows. These endpoints will connect with cloud providers like Azure and Google Cloud, as well as I Cubed’s own hardware infrastructure. Usage data will be transparently tracked, and revenues fairly distributed to model stakeholders, ensuring continuous incentives for both developers and infrastructure providers. In doing so, I Cubed closes the loop from open innovation to real-world impact — ensuring models not only live in code but also operate, earn, and evolve in decentralized environments.

Community building

A thriving decentralized ecosystem cannot exist without an active, empowered community. I Cubed is committed to cultivating a developer- and creator-first culture by launching a dedicated DAO pool to fund community-driven initiatives. This pool will support activities ranging from model benchmarking and educational content to open-source tooling and community-led curation, ensuring that contributors at all levels are recognized and rewarded.

Beyond token incentives, the DAO will govern long-term infrastructure and economic policies, allowing stakeholders to co-shape the platform’s evolution. Community members will have voting rights not only over resource allocation but also over how visibility, validation, and value flow within the modelverse.

Looking ahead, I Cubed will also invest in auxiliary tools to support model training, such as decentralized dataset libraries, model evaluation sandboxes, and reproducibility frameworks. These efforts aim to lower the barrier to entry for AI creators and foster an environment where innovation emerges from the bottom up, driven by shared purpose rather than corporate control.

In I Cubed, community is not an afterthought — it is the protocol.

Intelligence Cubed is Building an Open AI Future — And You’re Invited

At Intelligence Cubed (i³), we believe that intelligence is a public good, not a private asset. We are building the infrastructure to ensure that power over AI is distributed — not hoarded — and that the future of artificial intelligence belongs to the many, not the few. Our mission is to empower creators, communities, and consumers through a new model of AI ownership and governance.

The Problem: Centralized AI Is a Dead End

Let’s be honest: the current AI landscape is broken.

• Most models are controlled by a handful of corporations.
• Creators rarely see long-term rewards or sustainable income.
• Users overpay for generic tools, while high-quality models go unnoticed.

Even worse, innovation is stagnating. A few gatekeepers decide what gets built, who gets funded, and which models get distributed. This isn’t just inefficient — it’s undemocratic. Centralization restricts creativity, limits competition, and silences diverse contributions.

As AI becomes embedded in healthcare, finance, governance, education, and culture, leaving it in the hands of a few is a risk we can’t afford. We’re already seeing the consequences: biased algorithms, lack of transparency, and a widening innovation gap.

We need a system where intelligence is owned, governed, and evolved by everyone — not dictated by a select few in closed boardrooms.

The Solution: Decentralization + Democratization

i³ is creating a new kind of platform — one where AI development and deployment is:

🧠 Owned by Creators Through Initial Model Offerings (IMOs), developers tokenize their models, stake ownership, and earn recurring revenue based on real-world usage.

📢 Governed by the Community We use prediction markets and futarchy-inspired voting to surface high-quality models — based on impact, not branding or hype. The crowd decides what matters.

📊 Valued by Usage, Not Branding Every time a model is used, that activity is recorded. These usage signals determine value, visibility, and rewards. No more invisible hands — just transparent metrics.

💰 Incentivized for Innovation Our create-to-earn system ensures that improvements, remixing, and new training data are not only possible — but profitable.

🗳 Democratized by Design No more opaque algorithms or editorial bias. Every user has a vote. Every creator has a path to visibility and reward. Excellence is surfaced through community consensus.

This is what democratized AI looks like — a participatory ecosystem where models succeed on their merits, not their marketing budgets.

Why Now?

🧱 The tools are ready: open-source models, decentralized compute, smart contract infrastructure.

💡 The need is urgent: 68% of users say AI subscriptions don’t meet their needs, and creators struggle to monetize their work.

🌎 The stakes are global: AI is shaping every industry, from policy to art, and how we choose to build it will define who benefits.

Decentralized AI isn’t a futuristic fantasy. It’s a present opportunity. Tokenized ownership, usage-based revenue, and community governance make it possible today. And if we don’t act now, the status quo will calcify — leaving creativity, innovation, and access in the hands of the few.

The future of intelligence is up for grabs. We can choose centralization — or we can choose democracy.

Get Involved

If you’re:

• A builder tired of seeing your models buried on centralized platforms
• A researcher who wants to open-source without sacrificing ownership or future earnings
• A user who believes in transparency, fairness, and pay-per-use accessibility

Then Intelligence Cubed was built for you.

We are looking for creators, collaborators, and early community members to help grow the most open, equitable, and creator-first AI ecosystem ever built.

Whether you’re training LLMs, building agents, or curating data, there’s a place for you in the i³ Modelverse. Let’s decentralize AI, together.

Let’s prove that when intelligence is shared, everyone wins.

🔗 Learn more: https://icubed6.godaddysites.com

📢 Follow us on X: https://x.com/I3_Cubed

At Intelligence Cubed (i³), we believe AI should be open, decentralized, and accessible to everyone. That’s why we created the i³ Fellowship — a unique opportunity for researchers, engineers, and builders to get funded for developing the next wave of open-source, tokenized intelligence.

Whether you’re training your own model, fine-tuning a niche LLM, or experimenting with decentralized compute, this is your chance to gain support, visibility, and a path to long-term sustainability.

No Web3 experience? No problem. Even if you have no background in blockchain or crypto, you’re absolutely welcome. If you come from a machine learning background, we’ll support you every step of the way — including helping tokenize and deploy your model. That’s our job.

Why the i³ Fellowship Exists

Today’s AI ecosystem is dominated by closed systems and centralized platforms. Innovation is often stifled by paywalls, proprietary APIs, and a lack of creator incentives.

The i³ Fellowship aims to flip that script. We support independent builders, open-source contributors, and researchers who are paving a new path toward AGI — one powered by community, collaboration, and crypto-native infrastructure.

What You Get as a Fellow

Research Funding

• $5,000 — $100,000 for individuals
• $7,500 — $200,000 for teams
• Bonus credits based on progress and community impact

Recognition & Visibility

• Demo Days and Showcases to feature your work
• Hackathons with prize pools and partner support
• Connect with investors, institutions, and developers in the i³ ecosystem

Tools & Infrastructure

• Model hosting and monetization on the i³ Modelverse
• Benchmarking, usage analytics, and tokenized staking rewards

Legal & Advisory Resources

• Licensing templates, IP guidance, and DAO governance tools
• Expert support from Web3 and academic mentors

Fellowship Tracks

Explore any of these areas:

Decentralized Model Training

Use distributed compute and on-chain logic to train performant models.

Fine-Tuning & Modular AI

Customize and stack models for targeted use cases in domains like healthcare, robotics, or education.

Tokenomics & Rewards

Design smart, fair incentive systems that reward creators for training, remixing, and contributions.

Evaluation & Benchmarking

Create transparent systems for rating, curating, and improving models as a community.

Who Should Apply?

We’re looking for:

• PhD students, researchers, and open-source builders
• AI engineers and LLM tinkerers
• Web3-native devs with interest in tokenized infrastructure
• Interdisciplinary teams exploring AI + crypto innovation

And again — if you only have an ML or AI background, you’re more than welcome. We’ll handle the Web3 side of things with you.

If you’re excited about decentralized intelligence, we’re excited to back your work.

Fellowship Timeline

A 12-month program with four progressive phases:

• Months 0–3: Onboarding and infrastructure setup
• Months 3–6: Hackathon + Midterm Review
• Months 6–9: Deep development + ecosystem collabs
• Months 9–12: Final showcase + community integration

Apply Now

The process is simple:

• Submit your project proposal and background
• Outline your goals, budget, and timeline
• Share past experience or open-source work

🔗 Click here to apply

The Time Is Now

Open-source AI is booming. But builders still lack infrastructure, visibility, and support. The i³ Fellowship bridges that gap — empowering you to experiment, collaborate, and earn.

We’re here to help you build the decentralized AI future you want to see.

Great intelligence shouldn’t be centralized. Let’s open it up, together.

Stay Connected:

🌐 Website: https://icubed6.godaddysites.com

X: @I3_Cubed

The Problem: AI Innovation is Constrained

Artificial intelligence holds transformative potential across industries. Yet access to the tools needed to build and share high-quality models remains heavily restricted. Barriers like high compute costs, closed systems, and lack of incentives have made it difficult for independent developers and researchers to thrive.

Training an effective AI model is not only a technical challenge — it’s an economic one:

• A 7B parameter LLM can require 10,000 GPU hours (~$480K)
• Fine-tuning for specialized domains can cost tens of thousands more
• Even high-performing models often remain unseen and unrewarded

At the same time, creators rarely have access to platforms that enable them to earn from their work or meaningfully connect with users.

A New Era of AI Collaboration

Recent breakthroughs in efficiency have made high-performance model training more accessible than ever. Optimizations in CUDA, architecture search, and sub-2-bit quantization techniques are unlocking new possibilities for rapid, low-cost AI development.

We believe these technical shifts demand an equally innovative infrastructure — one that is open, decentralized, and community-powered.

At Intelligence Cubed (i³), we’re building that infrastructure. A platform where:

• Creators can tokenize and stake their models
• Users can discover and pay-per-use without recurring fees
• Communities govern what succeeds, how it’s used, and how it evolves

The result is a healthier, more equitable AI ecosystem — one where value flows to those who build it.

Introducing the i³ Modelverse

Intelligence Cubed is building a decentralized, community-driven platform for AI model development and usage, powered by blockchain technology for transparency and security. It offers users staking choices and pay-per-use access to AI models while rewarding contributors with tokens through a “create to earn” model. Governance is managed by the DAO to ensure trust and fairness for all participants.

For Model Creators

Unlike traditional Web2 ecosystems where creators labor under tech giants without proper attribution, our Web3-native infrastructure enables true ownership and assetization of AI creations. I Cubed allows developers to build autonomous, fully on-chain AI models with Proof of Intelligence validation — transforming the one-way “own-and-share-to-earn” paradigm into a creator-first “create-to-earn” economy.

For Community Users

Users gain access to high-performance models serving niche domains at minimal cost, draw inspiration from our open-source ecosystem, and stake models according to their expectation. We incentivize community members to experiment, remix creations, and vote with their usage, turning passive consumers into active ecosystem co-builders.

How It Works: Layer by Layer

Our architecture is fully modular and decentralized:

User Interface Layer (UI Layer)

The frontend layer enables users to browse, test, and stake AI models via a marketplace interface, integrated with multi-chain wallets for seamless transactions and asset management.

Smart Contract Layer

Core logic for tokenizing models, managing open source pools (51% stake thresholds), and automating dividends. Built on Ethereum frameworks (Hardhat/Anchor), it supports dynamic ownership rules and community governance.

AI Model Management Layer

Handles decentralized storage, version control, and model validation via off-chain compute networks. Ensures model integrity and enforces open-source licensing upon stake completion.

Off-Chain Compute & Data Layer

Provides scalable model inference APIs and data feeds (via Chainlink) to minimize on-chain costs, while leveraging decentralized compute networks for performance validation and slashing triggers.

Tokenomics Layer

Governs the dual-token system ($MOD for governance, utility tokens for rewards) and staking economics (dynamic APY, penalties). Uses Chainlink oracles to track usage data for automated payouts.

A Collaborative Ecosystem

We are building more than a platform — we are enabling a new model for AI collaboration:

• Developers retain ownership and earn from their models
• Researchers receive recognition and compensation for open-source contributions
• Users gain utility-driven access to tailored AI tools
• Contributors stake in innovation and share in its success

By aligning incentives across all participants, i³ fosters an ecosystem that is both economically and creatively sustainable.

What’s Next

Our MVP is live. Staking infrastructure is in place. We are partnering with leaders in academia, open-source development, and decentralized infrastructure.

We are creating the model layer for decentralized intelligence.

If you are a builder, researcher, investor, or curious explorer — join us.

The future of AI doesn’t belong to gatekeepers. It belongs to communities.