Bounded worker tests
Participating computers can take assigned work and return results for checking. That tests coordination, not large-scale learning.
Epistemic Cognition · Research project
Nagi
A small neural-network prototype exploring how learning work could be shared across participating computers.
Distributed training is still under development. We are seeking funding for the infrastructure needed to take the research further.
Discuss research fundingModelEarly experimental network
TrainingDistributed workflow in development
FundingInfrastructure support sought
Where it stands
Early experiments with a narrow model.
The current model takes three numeric inputs and makes a binary prediction. It can save checkpoints and run small training experiments. Larger training runs and broader evaluation are still ahead.
Measured acceptance
Work is accepted only when a defined check shows improvement. We still need stronger benchmarks and held-out tests to learn what carries forward.
Open questions
We have not demonstrated a generally capable model or a dependable training network at scale. Those are research goals, not current product claims.
What comes next
We need to build the training infrastructure.
Nagi needs more than a working prototype. We are seeking support for the compute, engineering, and independent checks required to find out whether this approach can produce useful learning.
Contact us about funding- 01 Compute and storage for repeatable training runs
- 02 Reliable work coordination across computers
- 03 Stronger evaluation and independent security review
Testnet records
See what has actually been tested.
These records show a permissioned Testnet job moving from funding to an accepted result, a reward claim, and a controlled swap. Testnet assets are not real-money payouts, and this work is not a public token launch.