Sentient Foundation is developing an open artificial intelligence ecosystem in which models, agents, data, and tools can be created and coordinated independently of a single company. The project grew out of the Sentient initiative, launched in 2024 at the intersection of AI, blockchain, and cryptography, while Sentient Foundation was introduced as a separate nonprofit organization in February 2026. Its central idea is Open AGI: an open AI infrastructure where developers can publish technologies, verify authorship, and receive economic incentives for their contributions. To support this vision, the ecosystem combines the GRID network, OML research, the ROMA multi-agent framework, and the SENT token.
Contents
- Sentient Foundation and the Open AGI Concept
- GRID, ROMA and Open AI Architecture
- OML, Loyal AI and Protection of Open AI Models
- SENT Token, Tokenomics and Sentient's Blockchain Economy
- Sentient Foundation Development and the Future of Open AGI
1. Sentient Foundation and the Open AGI Concept
Sentient's history began before the Foundation itself was established. In 2024, Sentient Labs raised $85 million in a seed round co-led by Founders Fund, Pantera Capital, and Framework Ventures. Key contributors to the project included Polygon co-founder Sandeep Nailwal, Princeton University professor Pramod Viswanath, and Indian Institute of Science researcher Himanshu Tyagi. The main objective was to develop open AI infrastructure with cryptoeconomic mechanisms for ownership and coordination.
Sentient Foundation was officially launched in February 2026 as a nonprofit organization serving as a neutral coordinator of the ecosystem. Its mandate covers research, governance, developer funding, and the advancement of open-source AI. The Foundation and Sentient Labs should not be treated as the same legal or functional entity: Labs focuses on technical development, while the Foundation is oriented toward governance and the growth of the open ecosystem.
The term Open AGI in Sentient's materials describes an approach to artificial general intelligence in which key technologies should not be controlled by a single corporation. It does not mean that a fully functional AGI has already been created. In practice, Sentient is currently developing tools, models, multi-agent systems, and coordination mechanisms intended to become components of a future open infrastructure for intelligent systems.
An important feature of the concept is the combination of open source with economic incentives. Publishing a model with open code or weights improves accessibility, but makes it more difficult for creators to control its use and monetize their work. Sentient aims to address this challenge through cryptography, blockchain, mechanisms for verifying the provenance of AI assets, and reward distribution among contributors.
2. GRID, ROMA and Open AI Architecture
A core component of Sentient's infrastructure is GRID, a network designed to connect different artificial intelligence components within a shared layer. Instead of sending every request to a single general-purpose model, the system can use a set of specialized resources. In the documentation, these components are called Artifacts and can include models, agents, data sources, tools, and computing services.
The GRID concept is based on composition. One AI component can search for information, another can perform specialized reasoning, a third can verify the result, while a separate tool can interact with an external service. This approach differs from the centralized chatbot model, where the primary intelligence and infrastructure are controlled by a single provider.
Sentient's architecture includes several key components:
- GRID — a network connecting models, agents, data, and tools;
- Artifacts — independent AI components available to other applications and agents;
- ROMA — an open framework for building and coordinating multi-agent systems;
- Open Deep Search — a search and reasoning system based on open models;
- OML — a research approach for open, monetizable, and governable AI models;
- Dobby — a family of open language models used for experiments with community-owned AI;
- SENT — a token for coordination, governance, payments, and incentives within the ecosystem.
ROMA, or Recursive Open Meta-Agent Framework, addresses a separate challenge: coordinating multiple AI agents. A complex request can be divided into subtasks, with different agents handling individual parts of the work before the results are aggregated and verified. This allows specialized models to be used instead of relying on a single large LLM.
Sentient already publishes part of its research infrastructure openly. Repositories, models, and technical research are available through GitHub, Hugging Face, and academic publications. However, the scope of the proposed GRID concept is broader than the individual components available today. The Open AGI architecture should therefore be distinguished from existing products: some technologies are operational, while the fully decentralized system remains under development.
3. OML, Loyal AI and Protection of Open AI Models
One of Sentient's most distinctive developments is OML, or Open, Monetizable and Loyal AI. Research on OML was published in 2024 and combines methods from artificial intelligence, cryptography, and blockchain. Its goal is to address the tension between keeping a model open and allowing its creators to retain control over its economic use.
The first practical implementation was OML 1.0 with fingerprinting technology. Hidden prompt-response pairs are embedded into a model through additional training and act as identifiers. The owner knows the secret prompt and can test a suspected copy of the model through an API: a characteristic response can provide evidence of the model's origin. Sentient released tools for this fingerprinting approach as an open project.
| Component | Function | Role in Open AGI | Current Format |
|---|---|---|---|
| GRID | Connects AI resources | Coordination of open intelligence | Developing network |
| ROMA | Agent orchestration | Multi-agent reasoning | Open-source framework |
| OML | Model protection and monetization | Ownership of open AI assets | Research and tools |
| Dobby | Language models | Experiments with community-owned AI | Open models |
| SENT | Incentives and governance | Economic coordination | Crypto token |
The broader Loyal AI concept proposes that a model should not only be open but also owned, controlled, and guided by the community that creates it. Sentient's research explores different implementation methods, including Trusted Execution Environments, cryptographic techniques, and AI-native cryptography. Fingerprinting represents one of the first practical approaches rather than a complete solution to every challenge associated with controlling open models.
As an experiment, Sentient released the Dobby family of models. In particular, Dobby-70B is based on Llama 3.3 70B Instruct and was further trained to develop its own behavioral characteristics. Dobby demonstrates how the project applies open-access and fingerprinting concepts to a real LLM. However, claims about community ownership should be understood in the context of an experimental governance model rather than as a universally established standard for AI ownership.
4. SENT Token, Tokenomics and Sentient's Blockchain Economy
SENT is the cryptoeconomic component of the ecosystem. Its tokenomics were detailed in January 2026. The maximum initial supply is set at 34,359,738,368 SENT, equivalent to 235. The Foundation describes SENT as a coordination layer for GRID, the network, governance, and reward systems designed to incentivize contributions to open AI development.
A total of 65.55% of the initial supply is allocated to the broader community. Of this amount, 44% is designated for community initiatives and an airdrop, 19.55% for the ecosystem and R&D, and another 2% for a public sale. The team receives 22%, while investors receive 12.45%. Team and investor allocations are subject to extended vesting periods, while parts of the community and ecosystem allocations become available earlier.
SENT utility includes staking, governance, fees, and payments. Staking is intended to enable participation in governance, direct funding toward AI initiatives, and provide access to selected Artifacts. Developers can also use their stake as part of participation in tasks and reward programs. Within Sentient DAO, staked tokens are connected to voting on emissions, treasury spending, and protocol changes.
The economic model also provides for payments between services. Agents, models, datasets, and other Artifacts can use SENT when interacting with each other, creating an on-chain settlement layer on top of the AI infrastructure. Annual emissions are set at up to 2% and directed to the Community Emission Pool, while any unused portion of the annual limit is intended to be locked.
When evaluating SENT, it is important to distinguish its proposed utility from actual demand for the token. The value of this economy depends on GRID adoption, the number of developers, payments between AI services, and governance activity. The token itself does not prove that AGI has been achieved or automatically make AI computation decentralized: blockchain primarily provides coordination, ownership, incentives, and settlement functions.
5. Sentient Foundation Development and the Future of Open AGI
In 2026, Sentient expanded its activities beyond developing its own technologies. In June, the Foundation announced a $42 million Open Source AGI Grant and Investment Program. It includes grants for researchers and open-source developers, as well as investments in companies building commercial products based on open AI.
Selection criteria include technical quality, ecosystem contribution, openness, and long-term potential. Projects do not need to publish their entire technology stack: it is sufficient for a meaningful part of the project to remain open. In addition to funding, the Foundation offers computing resources and engineering support.
This strategy complements GRID, which is designed to connect independent AI components. Grants and economic incentives are intended to expand the network of models, agents, data, and tools while allowing developers to verify the provenance of their technologies and receive rewards without depending on a single centralized platform.
Open AGI, however, remains a long-term objective rather than a completed artificial general intelligence system. Major challenges include scaling multi-agent systems, maintaining the quality of open models, computing costs, intellectual property protection, agent security, and the effectiveness of token-based incentives. Fingerprinting also does not solve every problem associated with copying or modifying AI models.
Sentient Foundation is therefore better understood as a blockchain-AI ecosystem for developing open intelligence rather than an already completed decentralized AGI. The project combines ROMA, OML, Dobby, and Open Deep Search with the broader GRID architecture and the SENT economy. Its future development will depend on whether this system can effectively coordinate independent AI developers while preserving openness and sustainable economic incentives.
