NEAR.ai → NEAR Protocol → NEAR AI
In 2017, a groundbreaking AI paper titled Attention Is All You Need was published, which made waves in the academic community and shaped the current AI landscape. Among the eight authors, seven went on to start their own entrepreneurship, while one later joined OpenAI. One of these authors is Illia Polosukhin, the co-founder of NEAR AI. Inspired by the paper and Illia’s entrepreneurial efforts in the blockchain industry, IOSG’s founding partner, Jocy, visited NEAR’s office in 2018. At the time, their office was located in an underground co-working space in San Francisco, where a group of engineers gathered to hold workshops. IOSG had long conversations with the founding team, helped Illia secure a visa to visit China, and assisted them in organizing over four offline developer events within just three months.
As a crypto OG project with a seven-year legacy, NEAR’s two co-founders are prominent figures in the AI field. This includes Illia Polosukhin, co-creator of the Transformers architecture, and Alexander Skidanov, an expert in MemSQL. Their initial vision was to leverage computer science students worldwide to train models and apply program synthesis to solve complex problems. As a result, NEAR was initially established as Near.ai driven by the belief that they could not only make groundbreaking advancements in program synthesis research but also develop a revolutionary product based on this technology. (Disclosure: IOSG participated in NEAR’s early investment rounds in 2018.)
In 2017, machine learning — especially natural language processing (NLP) — was still in its early stages compared to today, making the realization of program synthesis for practical applications a significant challenge. Their advisors recommended focusing on applying program synthesis to aid developers in programming verifiable smart contracts on Ethereum. However, Ethereum faced notable constraints at the time, particularly in scalability and microtransaction support, with no blockchain or traditional payment system capable of fully addressing these issues.
Therefore, NEAR spent six years building a blockchain that is both highly scalable and user-friendly — designed to support microtransaction needs and provide an easy onboarding process. It addressed challenges such as cold starts, key management, and account usability. However, this shift in focus caused the market to perceive NEAR more as a general-purpose Layer 1 blockchain with unclear use cases, rather than its original AI-driven vision.
As innovations in cryptocurrency and artificial intelligence continue to emerge, including decentralized compute marketplace, inference networks, and the rise of numerous AI agents memecoins, along with high-performing large cap AI tokens like TAO and WLD, NEAR’s pioneering role in this space seems to have been overlooked.
After seven years of development, NEAR has evolved into a scalable, developer-friendly blockchain ecosystem designed to support application builders. With its infrastructure now firmly established, Illia and Alexander are shifting NEAR’s focus back to its original vision: making NEAR AI a central platform for user owned AI.
During the IOSG OFR event in Dubai this April, Illia discussed his perspective on “Why AI Needs to Be Open” and emphasized his commitment to transforming NEAR into a leading ecosystem for next-generation AI research and applications.
Given the early stage of the Crypto x AI field, NEAR’s strategy focuses on advancing the Web3 AI ecosystem through nonprofit-driven initiatives. These include incubators, accelerators, investments, establishing open-source AI research centers and tools, and fostering community-driven AI development through competitions. The overarching goal is to ensure that AI advancements can be genuinely shared by the community. At this stage, the token value accrual for $NEAR is not their top priority so how these initiatives will contribute to the token economy has not yet been seen.
This article will delve into NEAR’s renewed focus on the Crypto x AI sector, helping readers better understand its positioning within the Web3 AI ecosystem.
Near AI x HZN Incubator
To echo this thesis, in May, Near foundation announced its launch of an NEAR AI x HZN incubator, an investment wing focusing on AI projects, as well as an R&D lab.
11 projects have been selected by the Near Foundation to receive support over the program’s initial twelve weeks, which ran from June to August 2024 and in subsequent phases through May 2025. That support includes access to NEAR’s AI expert network, dedicated technical and token launch support, access to compute, and funding opportunities.
These projects encompass various aspects of the AI stack, including data, models, computation, and inference. Some highlighted early projects milestones include:
- Hyperbolic, an open-access AI cloud, hosting the latest open source AI models, including Llama 3.1 405B
- Mizu launching their beta data pre-processing platform and attracting 20K users in the first week
- Pond launching their first GNN mode leveraging on-chain data for wallet predictions and achieving a 20% prediction rate
Near x Delphi Labs Accelerator
Since the earlier incubation has achieved great success, Near further teamed up with Delphi Labs, another active research and venture firm in the Crypto x AI space, to announce the first cohort of their joint AI Accelerator Program, a strategic initiative designed to support and rapidly scale high-potential projects building at the intersection of AI and Web3 running from October 15 through early December.
The selected teams will benefit from a comprehensive support package, including technical mentorship, funding opportunities, and access to NEAR and Delphi’s expert networks. Each team will receive a $100,000 investment from the NEAR Foundation, along with up to $250,000 in potential funding from Delphi Labs for further acceleration after the program concludes.
The program will culminate in a Delphi Labs demo day in mid-December, focused on showcasing projects to investors. Additionally, each project will receive $50,000 worth of computing credits from Aethir, the computing partner for the accelerator program. These resources will provide the computational capacity needed to develop and deploy high-resource AI models, enabling teams to focus on building their projects without being constrained by infrastructure limitations.
Unlike the incubator program where the focus is more on the foundational component of decentralized AI stacks, model, data and compute, the projects selected this time shed more light on the applications side.
- Almanak builds, tests, trains, optimizes, and deploys agents for DeFi
- Questflow aims to dominate the AI agentic workflow market, particularly in SEO
- Q Star Labs launched AI influencers that have organically grown to over 27,000 followers
Near AI R&D
NEAR itself officially dives back into the AI game with the release of Near AI on Nov 8 at its REDACTED event to position itself as the blockchain for AI along with three major developments.
Privacy Preserving AI Assistant
Near AI Assistant is a chatbot type of assistant that can answer any sort of generic questions for example, what is IOSG Ventures and Near AI? The AI Assistant has local memory for every user. It’s able to associate it with every NEAR account, remember things as the user chats to it and uses that memory when interacting with different agents.
But more importantly, it allows even non-technical creators and builders to tap into actions especially related to crypto. In just a few sentences, users can prompt the AI Assistant to create an image for memecoin, generate a website for the memecoin and launch it on Pump.Fun.
Open Source Toolsets for Building AI Agents
Near also released an Assistant API toolkit to help build, measure, and deploy AI agents and integrate it into specific applications.
Currently, there are 60+ agents uploaded to Near agents hub, with each focusing on diverse aspects of the Web 2 and Web 3 use cases, e.g. web scraping, token swap agent, youtube transcript agent. Each agent can connect and call on other uploaded AI Agents for task execution.
In future updates, Agents on NearAI can even transact with external services and APIs.
Access Community-built AI Research and Resources for Frontier Model Development
Near believes that open source contributions are key to unlock user-owned AI. To accomplish this goal, Near launched an AI Research Hub, where it provides all the toolsets and infra in one place for researchers to build open source AI: datasets, models, inference and decentralized storage + compute. A key feature among these open source toolsets is a benchmark tool that enables developers to compare different implementations of agents and models with the same functionality in the hopes to improve the metric over time.
To ensure SOTA model development, Near AI Research Hub leverages parallel competitions to continuously refine the models. An ongoing Model Training Series competition is happening right now to encourage the community to progressively train larger AI models from 0.5B parameters to 1.4T parameters against perplexity benchmarks, trained on FineWeb datasets and nodes with TEE and GPUs from the Core NEAR Protocol to ensure the model can be both private and monetizable. Check the leaderboard here.
Obviously, open source contributors need to be rewarded to capture the value of their creations and contributions. Future plans have been put on the roadmap to set up a rewards and royalties system for competitions and model usage.
Closing Thoughts
In NEAR’s vision of user-owned AI, everyone is encouraged to contribute, and every effort is valued and rewarded. Contributions aren’t limited to coding — activities like answering questions, helping others, and improving documentation are equally valuable.
As NEAR continues to expand its presence in the AI space, it has the potential to become the go-to platform for decentralized AI development, paving the way for the next generation of AI-driven dApps.
In fact, some AI-related dApps are already being built on NEAR. For example, a protocol called Bitte is focused on decentralized agent discovery, with use cases for its agent plugins including NFT minting and cross-chain transactions.
The NEAR ecosystem is also home to many other innovations beyond AI. To explore more projects, you can refer to a tweet shared by Marcus that highlights several exciting developments within the ecosystem.
Currently, the connection between NEAR’s AI strategy and its token economy or blockchain usage isn’t clearly defined. However, in the future, Web3 dApps developed on the NEAR AI platform could emerge within its ecosystem. NEAR tokens might be used to reward users participating in model training competitions or as payment for agent-based services. For instance, when users interact with NEAR’s agent services, they may need to pay with NEAR tokens.
Regardless of the specifics, this AI strategy represents a significant step toward encouraging more users to adopt decentralized Web3 AI, empowering them to take full control of their AI-driven futures.
NEARing User Owned AI was originally published in IOSG Ventures on Medium, where people are continuing the conversation by highlighting and responding to this story.
