
Author
Subarna Basnet
Published
Sep 2, 2025 • 1 min read
i found something crazy recently.
while exploring more things around AI infrastructure and how large models actually run, i came across a project called bittensor.
the idea immediately caught my attention.
right now most powerful AI systems are controlled by a few big companies because training large models requires massive computing power, huge datasets, and expensive GPU clusters.
but bittensor is trying something completely different.
instead of one company controlling intelligence, they are building a decentralized network for machine learning models. anyone can connect their models to the network, contribute useful outputs, and get rewarded if their model performs well.
basically a network where intelligence itself becomes part of the infrastructure.
they also have something called subnets, where different types of AI systems can run and specialize inside the network.
i'm still learning about how everything works -- validators, incentives, model competition, network coordination -- but the concept itself is really interesting. for now i keep reading the bittensor whitepaper, bittensor docs, and the subnet guide on understanding subnets.
for now i'm just exploring and trying to understand this ecosystem more.
this might be a very deep rabbit hole... and honestly i'm excited to keep exploring it.