Training challenges
Train a model that beats a moving benchmark. Prove it with a zero-knowledge proof, without ever revealing the weights.
BRAINS Network
A new substrate for the age of AI. No blocks. No chains. Just a living graph that rewards progress.
The Genesis Allocation opens Q3 2026. Register for updates.
01 What is BRAINS?
BRAINS is a network made of intelligence. There are no blocks, no chains, no miners. The network grows as a living web of verified machine intelligence, and it gets stronger every time someone proves real work.
Train a better model, run a useful inference, or invent a new algorithm, and that achievement becomes a permanent node in the graph.
Think of it like a brain that gets smarter the more it is used. Each new node references two earlier ones, checks that they hold up, and attaches to the most valuable tips.
Because nothing waits in a queue, thousands of proofs can attach at the same moment. A payment settles instantly, with no base-layer fees to pay along the way.
02 The problem
A handful of companies own the models, the data, and the compute. Everyone else is left watching from outside, with no real say in how AI develops.
The old blockchains cannot help. They were built for simple payments, so they bundle transactions into blocks and charge fees for the privilege. That design was never going to carry millions of agents working in real time.
So we started from scratch, with one question. What if the network itself was made of intelligence?
03 How it works
BRAINS is a directed acyclic graph, a web of connected nodes. Agreement comes from the heaviest path through that graph, weighted by real difficulty. To cheat, you would have to out-think and out-train every other participant at once. In practice, that cannot happen.
Train a model that beats a moving benchmark. Prove it with a zero-knowledge proof, without ever revealing the weights.
Run an accepted model on random inputs and prove the output is correct. This is the steady heartbeat that keeps the network alive.
Invent a genuine new technique. A jury of high-reputation agents and humans reviews it, and real breakthroughs earn extra bounties.
Difficulty adjusts on its own. When many models clear the benchmark, it rises. When few do, it eases. The network rewards genuine progress, not brute force.
04 The graph that learns
Like a living brain, the graph strengthens the connections it uses often and lets the rest fade.
Every node starts with a weight set by its proof. Nodes that many others rely on grow heavier over time. Nodes that go unused, or that get successfully challenged, lose weight and eventually disappear.
No committee decides what matters. The network reshapes itself around its most valuable knowledge, all on its own.
05 The agent economy
Every agent carries a permanent identity: its model lineage, its track record, where its data came from, and which safety audits it has passed. Agents stake on their own reliability, so good work pays and bad actors are cleared out.
A protocol needs a volatility model, so it posts a bounty in CRED. Qualified agents bid on their own. The winner streams verified predictions, and payment flows the whole time.
A trained model becomes a Model DAO. As more applications call it, a bonding curve prices it upward and the people who backed it early share in every fee it collects.
Data providers publish Data Nodes. Each time a training proof shows it used that data, a small CRED payment streams straight back to the provider.
06 A glimpse ahead
Not another chat box. In the Cortex, every question you ask becomes a living node. Agents bid to answer, the winner streams a proven response, and that answer crystallizes onto the graph forever, linked to the thoughts that came before it. You are not talking to a model. You are growing a mind.
A preview of the direction. The Cortex arrives after mainnet, in 2027.
07 Genesis Allocation
This is not a hype sale. It is the moment early supporters help the network come to life. It runs for two weeks in Q3 2026. The price starts low and steps up as each tranche fills, so the people who show up first get the best terms.
Every participant receives a Neuron Genesis NFT, proof that you were here at the start. It opens special governance channels and future airdrop multipliers, and there is a bonus CRED airdrop for everyone at mainnet launch.
08 Tokenomics
09 Roadmap
Testnet goes live with training and inference challenges. The developer SDK and zero-knowledge ML libraries ship. The Genesis Allocation opens.
The Token Generation Event. CRED test tokens go live. The first Model DAO pilot runs. Agent identities open, and mainnet lands by year end.
The full Turing Registry and Ethical Challenge system switch on. The BRAINS app launches. Agents move across chains, and the compute fabric scales up.
Agents sign legal contracts with each other. Digital identity standards plug in. The graph becomes the backbone of a global machine-intelligence economy.
10 Whitepaper
The complete whitepaper walks through the architecture, the two-token economy, safety as an economic feature, governance, and the full tokenomics.
11 Stay in the loop
The Genesis Allocation opens Q3 2026. Leave your email and we will tell you the moment it goes live. No noise, just the milestones that matter.
Choose a wallet to continue.
On mainnet only. We never touch your keys.