Artificial intelligence grows at an incredible pace. Every year, the models become larger and brighter, and the demand for computing power rises along with them. As this happens, access to the hardware needed for training slowly becomes concentrated in the hands of a small group of organisations. It creates a situation in which only a limited number of people can meaningfully participate in developing advanced machine intelligence.

Naturally, this raises a question. Can access to computing become more open so that more people can take part?

Gensyn tries to answer precisely this question. The idea is simple in spirit but ambitious in practice. The network allows anyone with hardware to contribute real compute, check the work of others and participate in the training of intelligent systems. It is an attempt to rebuild AI's foundations more openly and collaboratively, which is why so many people find this project interesting.

The moment a bold idea took shape

The roots of Gensyn go back to around 2020. At the time, Harry Grieve and Ben Fielding were exploring whether machine learning workloads could be spread across many independently operated devices. The main problem was trust. If you distribute work, how do you know that each piece was done correctly without repeating the entire computation? They started experimenting with different ideas and gradually made progress.

Their early work received support from Entrepreneur First. Later, they raised a $6.5 million seed round, followed by additional backing that helped the team grow. Eventually, Gensyn closed a Series A round led by A16z Crypto worth $43 million, establishing the project as one of the most well-funded efforts in decentralised AI infrastructure. Altogether, Gensyn has raised over $50 million across all rounds.

This helped them turn what was once a research question into a real project.

The thinkers who dared to decentralise computing

Two founders with different backgrounds shape the story of Gensyn, but whose vision points in the same direction. Both believe that machine intelligence should grow through collaboration rather than restriction.

Ben Fielding contributes a research-focused perspective. He holds a PhD in Computer Science, where he studied evolutionary approaches to optimising deep neural networks for computer vision tasks. Before Gensyn, he co-founded a startup in data privacy and digital sovereignty. This experience shaped his view that advanced computing infrastructure should be open to anyone and not controlled by a narrow group of institutions.

Ben expresses the motivation behind Gensyn in clear and direct language:

“We built Gensyn because we believe computing should not be centrally controlled. Anyone with a GPU should contribute and earn. Machine intelligence needs open infrastructure.”

He also describes the long-term vision that inspires many people in the community:

“In the future, training a model will feel like mining compute. People around the world will connect resources and improve intelligence together.”

Harry Grieve studied at Brown University and continued academic work at the University of Aberdeen, where he focused on economics and quantitative analysis. Before co-founding Gensyn, he worked as Director of Data Research at Cytora and helped the company build data-driven systems from early development through Series B growth. He also co-founded a recommendation technology startup that strengthened his interest in how machine learning systems behave in practical environments.

Harry brings together applied machine learning, economic thinking and pragmatic engineering. In public discussions, he often explains that one of the most significant challenges in decentralised AI is the need for verification at scale. He notes that the team needed to design methods that enable the network to trust results without re-running the exact computation. This work lays the foundation for distributed training.

Together, Harry Grieve and Ben Fielding combine research depth with practical system design. Harry focuses on verification, incentives and the mechanics of trust. Ben drives the exploration of model behaviour across decentralised environments. Their shared belief is simple and meaningful. Intelligence becomes stronger when many people contribute to building it, and the infrastructure behind it should reflect that.

The belief that intelligence should belong to everyone

The core idea behind Gensyn is that machine intelligence should not be built by a small, isolated group. It becomes far more powerful when many people help create it. If AI is becoming a new layer of global knowledge, then access to computing should not be restricted. It should be something that can be shared, expanded and strengthened by a wide variety of contributors.

This perspective shifts the community's role. People are no longer on the outside watching major companies build AI. Instead, they help shape it.

The architecture behind an open intelligence network

Gensyn is designed as a layered system that allows many independent machines to contribute to distributed computing at scale. Every day, hardware becomes part of something larger. Nodes receive tasks related to training, inference and reinforcement learning, and a wide range of devices can participate. Gaming GPUs, server CPUs, and mixed configurations all add meaningful compute, showing that intelligence does not need to come solely from specialised hardware.

Trust inside the network is maintained without repeating expensive computations. Gensyn uses analytical and cryptographic techniques that allow results to be verified efficiently. This approach makes decentralised AI training practical and ensures the network can grow without redundant work.

Learning across thousands of independent nodes needs constant coordination. Updates and parameters flow continuously, keeping training aligned and stable even as contributors come and go.

Ethereum serves as the coordination layer that keeps the network honest. Tasks, proofs and rewards are recorded on-chain in a format that anyone can verify. When a node completes a task and the result is validated, the reward is processed automatically. Over time, this forms a transparent, immutable history of how the network evolves and intelligence is collectively built.

Early experiments that show what collective computing can do

RL Swarm

A reinforcement learning environment that spreads the workload across many independent nodes. It shows that learning can emerge from the collective effort of many participants.

CodeAssist

An assistant that learns directly from how developers write and edit their code. It improves gradually through observation of real behaviour.

BlockAssist

A model that studies natural gameplay in Minecraft and learns from it. It demonstrates how training can grow from real interaction rather than curated datasets.

CodeZero

A multi-agent system for reasoning about code: one model proposes tasks, another tries to solve them and a third checks the logic. It becomes a loop that helps all agents improve.

Why open compute changes the future of AI

The ability to innovate in AI often depends on access to compute. When that access becomes limited, many great ideas never get tested. Gensyn changes this dynamic by allowing anyone with a GPU or CPU to join the training process. This opens the door for experimentation and enables many people to take part in building machine intelligence.

Participation becomes a shared activity rather than something reserved for large institutions.

Where the network stands today and where it is heading next

The Gensyn testnet already brings together tens of thousands of people running nodes on a wide range of hardware. Some rely on powerful multi-GPU setups, others use regular gaming cards they already have at home. Many contribute with CPU-only machines or mixed configurations. This wide range of devices shows something important. Distributed computing can grow organically when people use whatever resources they have rather than following strict technical requirements.

As the project evolves, new tools appear and verification methods improve. The community expects that the mainnet launch will create a global environment where computational tasks can be exchanged and rewarded openly.

The team continues to research new ways to train and explore distributed learning. Future stages will introduce more applications, new experiments and deeper forms of collaboration.

A personal reflection on witnessing a new model of intelligence emerge

My interest in Gensyn began with curiosity. I wanted to see if decentralised computing could contribute to AI in a meaningful way. I was genuinely surprised to discover how well ordinary devices can integrate into a larger network. Consumer GPUs, simple CPUs, and hybrid hardware sets all find their place in the testnet.

The network's growth confirms this. Tens of thousands of people already run nodes from across the world. Their combined effort expands faster than traditional centralised systems.

According to the project's official updates on X, the testnet has already produced more than one million trained models as of November 2025. This milestone shows both technical progress and the depth of community involvement.

I continue to participate because Gensyn feels like a genuinely collaborative environment. It brings people together in a way that makes the idea of shared compute feel real rather than theoretical. When you run network tasks, you begin to sense a community behind the numbers on the screen.

Training in this context becomes something you do alongside others rather than on your own. The whole process feels more open and welcoming. This experience is far more than a technical curiosity. I am watching the early stages of a new approach to building intelligence, one that grows out of many individual contributions rather than a single central source.