null Guests: Harry Grieve, Co-founder of Gensyn Edited by: momo, ChainCatcher
When the thirst for computing power of AI models meets the natural bottleneck of centralized supply, a computing power revolution is quietly taking place. Gensyn's two co-founders, Harry Grieve and Ben Fielding, have insightfully observed that the key to breaking the deadlock lies in activating the dormant computing power potential in billions of edge devices worldwide, and the path to this is decentralization.
Gensyn is committed to building a distributed machine learning network that connects idle computing devices worldwide through blockchain protocols, ensuring the reliability of training results with its innovative verifiable computing technology. Its testnet has attracted 150,000 users and is running stably. With the successful completion of the testnet phase, the Gensyn mainnet will also launch in the near future.
Gensyn has raised $43 million in Series A funding led by a16z, with a total funding amount exceeding $50 million. In this exclusive interview, Harry Grieve systematically explains how Gensyn builds the technical blueprint and business thinking for the next generation AI infrastructure starting from the core proposition of “breaking the scale.”
The purpose of decentralization is to break the limitations of computational power scale.
- ChainCatcher: Please introduce yourself first. What were the three most critical experiences before founding Gensyn? How did they shape your entry into the “decentralized AI computing” space?
Harry Grieve: I am part of the early generation that was exposed to the internet. Back then, the network was more open, decentralized, and filled with file-sharing networks and various information repositories. This shaped my understanding of information and the network, leading me to lean towards the ideals of open source and decentralization from an early age.
During my time at university and afterwards, I was exposed to classical liberal ideas, which made me pay more attention to individual rights and freedoms, and began to question centralization and censorship. This is directly related to today's AI models—when models make decisions for us, who decides their “rights” and behaviors? This prompted me to think about the relationship between AI, sovereignty, and ethics.
After graduation, I worked at a machine learning company in London and personally experienced the immense difficulties of accessing large-scale computing resources and high-quality data. I realized that to continuously develop more powerful models, the underlying resource (computation and data) access and scaling issues must be addressed, which is also the reason I later firmly entered the decentralized AI computing space and founded Gensyn.
- ChainCatcher: What was the impetus for the birth of Gensyn? How did you and Ben Fielding decide to go “All-in” on this direction from 0 to 1 in 8 weeks at Entrepreneur First?
Harry Grieve: We met at a social event before the UK accelerator program Entrepreneur First. The reason we were able to quickly decide on the “All-in” direction was based on two key consensus points:
First of all, we firmly believe that machine learning is the future. In 2020 (before the emergence of ChatGPT), we were all highly convinced that machine learning would be the next technological wave. Although this was not a consensus at the time, we witnessed technological breakthroughs in areas such as image generation and interaction, and we deeply believed in its potential.
Secondly, we both oppose “centralization.” I am constrained by the bottlenecks of centralized computing and data sources, while Ben focuses on personal privacy and data security in his doctoral research and entrepreneurship. We are both critical of centralization. Initially, we were concerned with technologies such as “federated learning,” but later realized that to solve the trust issues, a decentralized state recording and accountability mechanism is needed, which ultimately led us to blockchain. We transitioned from being “AI native” founders to “AI + crypto” explorers.
- ChainCatcher: At that time, AI training seemed to mainly rely on cloud giants (like AWS). Why do you firmly believe that “decentralized computing power” will definitely have opportunities? What do you think was the main driving force behind choosing decentralized computing power at that time?
Harry Grieve: The driving factors are multifaceted, but the core answer is scale.
Most of the available internet data has been used to train models. Future performance improvements rely on accessing data that is located at the “edge” and is currently inaccessible. To make use of this data, you must move towards the edge, which inherently requires decentralization.
Despite the huge investment in centralized computing power, the demand for computing power from AI is “endless.” This thirst will drive the demand for computing power to spread to all underutilized devices. The only way to connect and scale these decentralized resources without centralizing them all is through decentralization.
So, scale is the only answer. Decentralization is to unlock an unprecedented scale of computing and data resources.
What is the core differentiation of Gensyn?
- ChainCatcher: If I had to explain in one sentence to an audience without a technical background, what kind of system is Gensyn “building”?
Harry Grieve: Gensyn is a system that allows you to access all the core resources needed to build machine learning systems (such as computing power and data) on an unprecedented scale.
- ChainCatcher: The decentralized computing track already has players like Akash, Render, io.net, what is Gensyn's focus or core differentiation approach?
Harry Grieve: We have great respect for early players like Akash. Our core differentiation lies in the different perspective on resources: other projects primarily offer singular, containerized GPU computing power leasing. In contrast, Gensyn has a broader perspective, considering multiple machine learning resources (computing power, data, models), which are interwoven and recyclable.
For example, the output generated by a node performing model inference is data, and this data can be used to train other models. In our network, the boundaries between inference, training, computation, and data become blurred. The system we built is designed to adapt to this dynamic, chaotic new paradigm of machine learning.
- ChainCatcher: Can you explain to the readers about the current product layout that Gensyn is working on? How are you systematically achieving decentralized computing power?
Harry Grieve: This is a technical description: it is a decentralized cryptographic network where users can access various resources through our native token—whether it is verifiable computing resources for training or inference, or mechanisms that incentivize different model training by setting objective criteria. This system comprises three core components that together form a powerful closed loop:
Verification System: This is our core technology. We have developed a proprietary compiler and verification framework capable of achieving bit-level precise verification across different hardware and software. This means we can prove that the training results of a model on one device are completely consistent with the results verified on another completely different device. This is the cornerstone of building network trust and preventing fraud.
Extension Technology (Swarm): This is a peer-to-peer training framework (such as for human feedback reinforcement learning). It allows you to connect countless devices worldwide for horizontal scaling, utilizing computation and data on edge devices for training, thereby creating more powerful models.
Assist Agent: We have autonomous AI assistants that can be integrated into applications. They can learn in an unguided manner and assist users in completing tasks. When these assistants are being trained, they can leverage our extended technology for cross-device training, allowing them to self-evolve and become stronger.
Overall, when users integrate intelligent assistants into applications, they continuously generate interaction data during the execution of tasks; subsequently, this data is input into our extended technology framework, where it continuously optimizes the model through a cross-device collaborative distributed training method. During this process, core validation technologies ensure the accuracy and reliability of the training process, ultimately producing a new generation of models with significantly enhanced performance. This process forms a nonlinear, continuously reinforced machine learning ecosystem, allowing the system to maintain reliability and evolutionary capability while scaling.
- ChainCatcher: What is the biggest technological milestone achieved by Gensyn from the Series A funding in 2023 to the public testnet in 2025? Was there a moment that made the team “scream collectively”? What do you think is the most underestimated technological innovation of Gensyn at present?
Harry Grieve: To be frank, we probably scream more out of “fear” than out of “excitement”; entrepreneurship is tough.
I believe the most underrated technological innovation is actually our verification system. The construction of this technology is extremely complex, requiring a comprehensive solution to all potential factors that could lead to non-determinism, from compilers and machine learning frameworks to hardware layers (even including GPU bit flips caused by cosmic rays). There is a huge gap between its value and public awareness. It is this technology that ensures the security and scalability of our network, allowing us to confidently permit any device to join the network and perform verification without worrying about security being diluted.
Over 150,000 users on the testnet, the mainnet is about to launch.
- ChainCatcher: Compared to centralized cloud computing giants or other decentralized computing networks, do you currently have some advantages in performance efficiency and cost?
Harry Grieve: At absolute cluster scale, we cannot currently compete with giants like AWS, but this is primarily an issue of network adoption rather than technical limitations. Our advantage lies in unlocking new resource scales (especially in edge computing and data) and becoming the infrastructure for future machine intelligence civilizations. We believe that truly autonomous AI, capable of self-evolution and existing within the crypto-economic system, will require a decentralized, permissionless network as its “habitat,” which is exactly what we are committed to building.
- ChainCatcher: What is your current network activity level? What data is worth sharing?
Harry Grieve: During the testnet phase, we have made very positive progress: we have over 150,000 users, most of whom have grown naturally through the appeal of the product; about 40,000 nodes are running on the network; the system has trained over 800,000 models.
- ChainCatcher: What are the “last mile” obstacles to the mainnet launch? What is the mainnet timeline you set for the team? Is there a clear timeline for the TGE?
Harry Grieve: The launch of the mainnet is currently the top priority, and the TGE will follow. We are about 3-4 weeks away from the mainnet launch, after which we will begin the mainnet audit.
Before this, the main focus was to ensure that all mechanisms were in place, functioning correctly, fully operational, and most importantly, to guarantee that the economic activities of the network were secure.
- ChainCatcher: How has the market demand faced by Gensyn changed compared to its inception? What impact do you think the arrival of the era of machine intelligence will have on you?
Harry Grieve: Compared to the early days of its founding, the market environment that Gensyn faces has undergone a fundamental change. Looking back to when we first started in 2020, we still needed to repeatedly explain the importance of machine learning to investors. However, with the advent of ChatGPT, AI has become a consensus across society. This shift in perception has also brought about a more intense market competition, with various AI and computing startups emerging like mushrooms after rain. At the same time, the focus of industry discussions has changed significantly—issues such as the ethical boundaries of open-source models and the regulatory framework for AI governance, which were hardly mentioned a few years ago, have now become hot topics in policy-making across countries.
In this context, the accelerated arrival of the era of machine intelligence precisely confirms the value of Gensyn's existence. The decentralized computing network we are building is essentially meant to provide the underlying support for the upcoming autonomous evolutionary machine intelligence. When AI systems need to break through the existing computational power bottleneck to achieve true autonomous learning and rapid iteration, the infrastructure we have built will become a key cornerstone of this new era.
- ChainCatcher: You mentioned in your public speech the “economic, ethical, and regulatory challenges of AI.” What regulatory risks are you most concerned about? How does Gensyn's protocol design achieve a balance between “compliance-friendly” and “anti-censorship”?
Harry Grieve: When discussing the topic of AI regulation, my biggest concern is that regulatory policies may mistakenly target the infrastructure layer. Imagine if future policies were to limit the number of GPUs, the scale of datasets, or even impose restrictions on the proportion of electricity used for AI training; such a crude regulatory approach would severely hinder progress in the entire technology field. From our perspective, AI models should essentially be open-source and shared like mathematical formulas, and should not be overly restricted.
At the protocol design level, we are exploring a path to balance. The model weights and data transmission in the current network are still primarily in plaintext, which provides the necessary transparency for compliance regulation. At the same time, since we are built on underlying public chains such as Ethereum, we naturally inherit their characteristics of decentralization and verification mechanisms. This architecture maintains the necessary regulatory visibility while ensuring the system's resistance to censorship.
As AI capabilities continue to break through, finding a balance between openness and control will become an important issue that we and the entire industry need to continually face in the coming years.
- ChainCatcher: What are the key metrics of Gensyn's success when looking back from 2030?
Harry Grieve: The key indicator of success is not simply financial data or user numbers. I hope Gensyn's greatest contribution will be to serve as the economic foundation of a parallel machine civilization.
By 2030, I hope to see a fully parallel society, civilization, and economy operating on the blockchain, with no humans involved. This civilization can generate economic output comparable to or even greater than that of humans, possess true creativity, and significantly advance scientific development and solve major issues facing humanity (such as extending lifespan and reducing inequality). If Gensyn is the cornerstone for all this to be realized, it will be the ultimate mark of our success.
