Getting started
Proof of Neural Work
Brains
Security
Getting started
Introduction
Neurite is a store of value secured by Proof of Neural Work (PoNW). Instead of guessing hashes, miners run simulations of real brain connectomes — complete wiring diagrams of real brains, reconstructed from scans.
The protocol assigns each miner a fresh challenge. The miner's worker simulates the brain, and verified results earn mining shares. Every round, a defined amount of Neurite (NRT) is emitted and split between miners in proportion to their accepted shares.
- Work with structure
- Each job simulates a real neural circuit, not an arbitrary puzzle.
- Fresh, single-use jobs
- Every challenge is tied to one wallet and one round.
- Verified before rewarded
- Shares come only from computation the protocol accepts.
- Pro-rata emissions
- No lottery: each round is split by accepted work.
Note
Quickstart
From zero to your first verified share:
- 1
Connect your wallet
Connect your wallet from the Neurite dashboard. Jobs — and the rewards they earn — are bound to this address. - 2
Choose where the brain runs
Run the worker on your own device, or use a hosted worker if you'd rather not keep a machine online. Either way, rewards go to your wallet. - 3
Start a brain
Pick a supported brain model and start mining. The worker receives its first challenge from the protocol. - 4
Watch the dashboard
Follow the neurons firing, the spike raster, step progress, and each result as it is submitted and verified. - 5
Earn shares, receive emissions
Each verified job earns mining shares for the current round. When the round's emission is released, it is split across all accepted shares.
Tip
Core concepts
- Brain model
- A connectome-based network — neurons and the synapses between them — that miners simulate.
- Challenge (job)
- A unit of work: model, starting conditions, sensory inputs and step count, bound to a wallet and round.
- Simulation
- Running the brain forward step by step as inputs propagate through its connections.
- Brain state
- The result of a job: the network's resulting state after the final step.
- Verification
- The protocol's check that a result is valid for the challenge it answers.
- Share
- Credit for accepted computation within a round.
- Round
- A period of mining with its own defined emission.
- Emission
- The NRT released each round and split across accepted shares.
Proof of Neural Work
Challenges
Mining starts with a challenge issued by the protocol. A challenge spells out everything the simulation needs — and who it is for.
| Field | What it specifies |
|---|---|
round | The round the job belongs to. Its result only counts toward this round. |
wallet | The miner the job was issued to. |
model | Which brain model (connectome) to simulate. |
initial_state | The starting conditions of the network. |
stimulus | The sensory inputs applied during the run. |
steps | How many simulation steps must be completed. |
{
"round": 1284,
"wallet": "0x7a3f…9c21",
"model": "fly/whole-brain",
"initial_state": "0x5e1d…a09b",
"stimulus": {
"channel": "visual",
"pattern": "looming",
"seed": 88412
},
"steps": 10000
}Because the wallet and round are part of the job, a result can't be replayed in a later round or claimed by a different address. Old results simply don't answer any open challenge.
Simulation
The neural simulation is the work. Starting from the initial state, the worker advances the brain one time step at a time. Sensory neurons receive the stimulus; every neuron integrates the signals arriving over its connections; neurons that cross their threshold fire, sending spikes on to the neurons they connect to.
After the specified number of steps, the job has produced firing patterns and a resulting brain state. Completing that simulation is the mining task.
# Illustrative: one job, as a leaky integrate-and-fire simulation
for step in 1..steps:
for each neuron in brain:
neuron.v = neuron.v * leak # potential decays
+ synaptic_input(neuron) # spikes arriving over connections
+ stimulus(neuron, step) # sensory input, if this is a sensory neuron
if neuron.v >= threshold:
fire(neuron) # spike travels along its synapses
neuron.v = reset
result = (firing_pattern, final_brain_state)Note
Verification
Submitting a result earns nothing on its own. The protocol first checks it against the challenge it claims to answer — only accepted results become shares.
- Right job
- The result answers a challenge issued to this wallet, for the current round.
- Right work
- The submitted brain state is consistent with simulating the specified model, starting conditions, inputs and steps.
- Counted once
- Duplicate and stale submissions are rejected.
- Activity isn't evidence
- Self-reported spike counts are never used to decide rewards.
| Submission | Outcome |
|---|---|
| Valid result for your current job | Accepted — earns shares |
| Result from a previous round | Rejected |
| Result for another wallet's job | Rejected |
| The same result submitted twice | Rejected |
| An inflated spike count | Ignored — activity isn't rewarded |
Emissions
Mining rewards come from scheduled emissions: each round releases a defined amount of Neurite, split according to accepted work. Supply grows on a schedule, not with how much activity miners report.
Schedule
Brains
Brain models
A brain model is a connectome turned into a simulable network: every neuron and every synapse from the scan, wired exactly as mapped.
| Model | Scale | Notes |
|---|---|---|
| Fruit fly Drosophila melanogaster | ~140K neurons · ~50M synapses | Whole adult brain. Sensory inputs include visual (optic lobes) and olfactory (antennal lobes) channels. |
| Rat Rattus norvegicus | Published at release | Mammalian model built from rat brain scans; larger circuits and heavier jobs. |
Data sources
Connectomes are reconstructed from electron-microscopy images taken at nanometre resolution. Machine learning traces each neuron through thousands of image slices, and human proofreaders correct the result — which is what made whole-brain maps practical.
In 2024 the FlyWire consortium, led by Princeton University, published a complete wiring diagram of an adult fruit fly brain, reconstructed with the help of AI segmentation that includes work from Google Research. Neurite's fly model is based on this kind of public connectome data. The rat model's source and specifications will be documented at release.
Independence
Mining
Device & hosted workers
A worker is what actually runs the brain. You can choose either kind:
- Your device
- Runs the brain model on your own hardware. You control the machine; it mines while the worker runs.
- Hosted worker
- Runs on your behalf, so there's no machine to keep online. Jobs are still bound to your wallet, so rewards still go to you.
Heavier brain models need more compute per job. Requirements are published alongside each model.
Reading the dashboard
The dashboard shows your brain at work. You can try a live in-browser preview on the home page.
| Panel | What it shows |
|---|---|
| Status | Receiving job → Simulating → Submitting → Verified, for the current challenge. |
| Neural activity | The brain model with neurons lighting up as they fire and signals spreading along connections. |
| Spike raster | One row per neuron, one tick per spike, scrolling over time — waves of activity are easy to spot. |
| Simulation steps | Progress through the job's required step count. |
| Shares | Jobs that have been verified and credited this round. |
| Last result | The brain state committed by your most recent submission, and its verification status. |
Tip
Security
Anti-gaming design
If activity were the reward, the cheapest strategy would be to fake it. PoNW is designed so that the only way to earn is to do the work.
| Attempt | Why it doesn't pay |
|---|---|
| Spamming spikes | Activity isn't rewarded; only accepted computation earns shares. |
| Fabricating a counter | Self-reported numbers are never trusted — results are checked against the challenge. |
| Replaying an old result | Jobs are bound to a round, so stale results don't answer any open challenge. |
| Taking another miner's result | Jobs are bound to a wallet, so someone else's result doesn't count for you. |
| Splitting across many wallets | Rewards are linear in accepted shares — splitting earns the same total as one wallet. |
Reference
FAQ
What is Proof of Neural Work?
Which brains can I mine with?
Do I need special hardware?
Can I earn more by reporting more spikes?
Can I resubmit an old result?
How are rewards calculated?
Where do the brain scans come from?
Where can I follow updates?
Glossary
- Connectome
- A complete map of the neurons in a brain and the synapses that connect them.
- Neuron
- A cell that integrates incoming signals and fires when its potential crosses a threshold.
- Synapse
- A connection through which one neuron's spike influences another.
- Spike
- A neuron firing. Shown on the dashboard; not a reward metric.
- Brain state
- The condition of every neuron in the network at a point in the simulation.
- Challenge / job
- Protocol-issued work bound to a wallet and round.
- Round
- A mining period with a defined emission.
- Share
- Credit for accepted computation within a round.
- Emission
- Scheduled NRT released each round, split pro-rata by shares.
- Worker
- The software — on your device or hosted — that runs the brain simulation.
- PoNW
- Proof of Neural Work: Neurite's consensus work of verified brain simulation.
Community
Announcements, round updates and releases are posted on X.
@NRTbrainFollow Neurite on X