Gensyn
In 2020, two London founders bet that AI compute sat in too few hands.
Verdict: Gensyn spent five years building open AI compute, then rebuilt its homepage around "machines that predict the future."
A Narrative Authenticity Framework analysis of Gensyn. NAF score 7.5 out of 10. Weakest component: Real Aspiration, 6.5.
Originally published on LinkedIn, April 28, 2026, with a score of 8.3. Facts checked and updated November 2026.
In 2020, Harry Grieve and Ben Fielding met through Entrepreneur First in London and made a bet: the computing power needed to train AI was controlled by too few companies, and someone should build an open alternative.
Fielding knew the problem firsthand. He told TechFlow he did his PhD research with a few GPUs in workstations under his desk, while big tech companies ran similar experiments on thousands. The founders spent five years on the hard part: proving that strangers' machines had done the work correctly.
On April 22, 2026, Gensyn launched its mainnet. The first app on it wasn't a compute marketplace. It was Delphi, a platform for markets on future events that AI models settle.
The homepage changed to match. It now reads: "Our mission is to build machines that predict the future." The open-compute story that got Gensyn here isn't on it.
Where Gensyn stands
The founders. Ben Fielding, co-founder and CEO, and Harry Grieve, co-founder.
The money. More than $80 million raised from investors including a16z crypto, which led the 2023 Series A.
The year. Mainnet and Delphi went live on April 22, 2026. The $AI token launched on April 29 and listed on major exchanges. In September it released open-1b, which it called "the first model you don't have to trust," trained in a way anyone can audit.
The brand line. "Machines that predict the future," with Gensyn described as "a machine learning research and production lab."
The NAF scorecard
| Component | Score | One-line read |
|---|---|---|
| Friction | 8.0 | A real problem, named in news posts, not on the homepage |
| Lived Stake | 8.0 | Founders who hit the compute wall, now off the front page |
| Distinctive Qualification | 7.5 | A live network, still early in proving demand |
| Real Aspiration | 6.5 | "Predict the future" names a capability, not a changed world |
| Composite | 7.5 |
Scores as of publication. Current standings at seekargus.com/leaderboard.
Score change: Gensyn drops from 8.3 to 7.5. In April it was scored on a clear open-compute story. Since then the brand has moved to prediction, and the homepage no longer names the problem the founders set out to fix. Friction moves from 9.0 to 8.0, Lived Stake from 8.5 to 8.0, and Real Aspiration from 8.0 to 6.5. Distinctive Qualification holds at 7.5.
Friction: 8.0
The problem Gensyn was built for is still real: training AI takes huge amounts of compute, and a few cloud companies control most of it. Its open-1b release names a sharper version, that people are asked to trust AI models they can't inspect.
The homepage names neither. A visitor learns what Gensyn wants to build, not what's broken.
Lived Stake: 8.0
The founders came to the problem through their own work, and they've stuck with a hard technical bet for six years. That counts.
The new homepage doesn't mention them or why they started, and the move to prediction markets puts distance between the brand and the problem they lived.
Distinctive Qualification: 7.5
Gensyn has done something few rivals have: a live network that checks machine learning work without a central referee, plus an auditable model to show for it.
It's early. Analysts note that real paying demand is still less clear than token trading volume, and rivals such as Akash, io.net, Prime Intellect and Bittensor chase the same idea.
Real Aspiration: 6.5
In 2023, Gensyn described its goal as compute on the scale of owning your own cluster, with fair access that no central authority could control or block. That was a changed world worth building.
"Machines that predict the future" describes what the technology does. A 9 would say who benefits: AI that no single company controls, trained on compute anyone can join, and models anyone can check.
The story Gensyn should tell
In 2020, we saw that a few companies controlled the computing power AI runs on. We spent five years building a network where anyone's machine can do the work and anyone can check it. Now the models trained on it don't have to be taken on trust.
Four moves to raise the score
1. Say what predicting the future is for (Real Aspiration 6.5 to 8.0)
Under the homepage mission, add one line connecting it to the founding bet: AI that no single company owns, trained on an open network and checkable by anyone. Measure it by how often press and developers repeat it.
Prior-work check: Partial. Gensyn's earlier homepage, still online as an archive page, led with open compute, and blog posts such as "GPT@home" make the case. The current homepage doesn't, as of November 19, 2026.
2. Put "a model you don't have to trust" on the front page (Friction 8.0 to 9.0)
Lead with the trust problem open-1b answers: people are asked to rely on AI models they can't audit. Measure it by homepage engagement and developer sign-ups.
Prior-work check: Partial. The open-1b announcement makes this argument on Gensyn's news page. It isn't on the homepage, as of November 19, 2026.
3. Tell the founders' story on the site (Lived Stake 8.0 to 8.5)
Add a short page on why Grieve and Fielding started: the compute they couldn't get and the six years of work since. Measure it by time on page and press pickup.
Prior-work check: Partial. The founders tell the story in interviews and podcasts. The homepage names neither founder, as of November 19, 2026.
4. Publish real usage every month (Distinctive Qualification 7.5 to 8.0)
Report verified compute jobs, paying users and Delphi activity in one public monthly dashboard, separate from token figures. Measure it by developer adoption and analyst coverage.
Prior-work check: Partial. Binance Research says Gensyn's protocol documentation and live app data are public in its docs. No plain-language monthly usage report found, as of November 19, 2026.
| Move | Closest existing program | Status |
|---|---|---|
| What predicting the future is for | Archived homepage; "GPT@home" post | Partial |
| "A model you don't have to trust" up front | open-1b announcement | Partial |
| Founders' story on the site | Interviews and podcasts | Partial |
| Monthly usage report | Public docs and app data | Partial |
Projected composite: 8.4 (Friction 9.0, Lived Stake 8.5, Distinctive Qualification 8.0, Real Aspiration 8.0). These checks cover public sources only and can't see Gensyn's internal plans.
Sources
- Gensyn homepage, viewed November 19, 2026
- Gensyn news: open-1b and Delphi, 2026
- Gensyn's earlier homepage (archive page)
- The Block: Gensyn launches Delphi, April 22, 2026
- Binance Research: Gensyn mainnet and token timeline, 2026
- Runtime Wire: open-1b and Fielding's research origins, September 2026
- ChainCatcher: Epicenter interview with the founders on fair access, 2023
- Yellow: analysis of Gensyn's network and demand, 2026
Tim Donovan · SeekArgus