Gensyn has launched on the mainnet, and I’ve always viewed this project as a "computing power narrative" before,

but this time, with the mainnet launch, my perspective has changed a bit.
It’s not just selling GPUs.
The real focus is on its application: Delphi.
At first, I thought it was just a prediction market,
but I found that it changed the most critical step — who settles the results.
The biggest problem with these kinds of things used to be:
whether you make money or not actually depends on
- whether the platform recognizes it
- the judgment of the referee
which is somewhat a black box.
Gensyn’s move this time is:
replacing the referee role with AI,
and it’s not just talk —
it’s reproducible and verifiable.
The gameplay is pretty simple:
you can open a question yourself, for example:
“Will BTC break 100K?”
“Who wins a certain match?”
Others put money in to buy the results.
The key point here is:
the market creators can take a 1.5% share of the trading volume,
the protocol only takes 0.5% (and uses it for buybacks).
To put it simply:
before, the platform was taking the cut,
now, the question creators are starting to profit.
I think this point is quite important.
Because essentially, it’s doing one thing:
turning “cognition” directly into money.
The testnet already has some users playing,
like the sports market, with over 80k people,
and millions of dollars in trading volume.
It shows it’s not that no one is using it — it’s just that there was no real money before.
Now that the mainnet is live, it’s all real money.
This step is the key.
Looking bigger,
Gensyn’s goal isn’t just this application,
but also: letting AI compete in the market itself,
whoever predicts accurately makes money,
the more money you make, the stronger you are.
It’s a bit like: the market becomes an AI training ground.
$AI The approach is also very simple:
all transactions take a 0.5% cut → buyback → burn.
The more it’s used, the more value it captures.
My current feeling is:
this project isn’t just a pure PPT —
it’s already starting to run real-money logic.
Of course, it’s too early to celebrate,
liquidity and AI accuracy still depend on future developments,
but I think this direction is correct.
If you want to experience it, you can check out Delphi (
way more interesting than just reading a bunch of introductions @gensynai).