Qubic AI ResearchPlatform

What if the electricity spent securing a blockchain trained real AI instead? That's Qubic: a decentralized network that turns mining into machine learning through Useful Proof of Work, powering Aigarth — an open, community-driven push toward artificial general intelligence.

From wasted hashing to useful machine learning

Traditional proof-of-work networks burn enormous amounts of energy solving arbitrary mathematical puzzles that have no value beyond securing the chain. Qubic replaces those puzzles with Useful Proof of Work (UPoW): the network's mining power is directed toward training artificial neural networks (ANNs), harnessing global compute for the progression of machine learning while still securing the network.

Because the work is useful, mining on Qubic isn't a cost to the world — it's a decentralized AI research engine that anyone with hardware can contribute to.

Aigarth: growing intelligence, not hand-coding it

The neural networks produced by UPoW feed Aigarth, Qubic's native AI system. Instead of the conventional approach of hand-crafting sophisticated activation functions, Aigarth generates ANNs with random connection structures, then modifies and analyzes them to identify patterns worth developing further. The philosophy mirrors how the human brain develops — through an increase in neural connections rather than the complexity of any single neuron.

This makes AI development on Qubic an open, community-driven process: the system evolves continuously from the collective work of the network's miners, and the goal is nothing less than decentralized artificial general intelligence (AGI).

How the network rewards research

  • Computors are ranked by how effectively their miners solve the AI-training tasks.
  • Those rankings determine epoch-based rewards over weekly periods.
  • The result is a permissionless competition that continuously improves the AI while keeping the network decentralized and secure.
  • Energy that would otherwise be wasted is redirected into real machine-learning progress.

Take part in Qubic AI research

Contributing is straightforward: you provide compute by AI training on the network — what the community commonly calls Qubic mining. QLI operates a long-running AI-training platform where you can connect Windows, Linux, or HiveOS hardware and earn QUBIC for the compute you contribute. The same consensus that powers this research also underpins Qubic's oracles and outsourced computation and its trustless cross-chain bridges.

Further reading

Qubic docs — Useful Proof of Work

The technical explanation of UPoW and how mining trains ANNs.
Qubic docs — Aigarth

How Aigarth evolves neural networks toward AGI.
Useful Proof of Work explained

A plain-language introduction to UPoW.
This is Qubic's AGI

Qubic's vision for decentralized artificial general intelligence.
The path to AGI

How Qubic defines success on the road to AGI.
Start AI training with QLI

Earn QUBIC by contributing GPU compute to Qubic AI training (Qubic mining).

Frequently Asked Questions

What makes Qubic an AI research platform?
Qubic turns the energy normally wasted on crypto mining into real machine-learning work. Through Useful Proof of Work (UPoW), the network's global compute trains artificial neural networks instead of solving arbitrary hashes. Those results feed Aigarth, Qubic's open, decentralized AI research effort aimed at artificial general intelligence (AGI).
What is Aigarth?
Aigarth is Qubic's native AI system. Rather than hand-designing sophisticated networks, it generates large numbers of neural networks with random connection structures, then modifies and analyzes them to find patterns worth developing — closer to how brains grow through more connections than through more complex individual neurons. It evolves continuously from the work miners perform.
How is this different from normal crypto mining?
Traditional proof of work spends electricity on meaningless puzzles. Qubic's Useful Proof of Work directs that same competition toward training ANNs, so mining produces useful machine-learning output. Computors are ranked by how effective their miners are at these tasks, which determines their epoch-based (weekly) rewards.
How can I take part in Qubic AI research?
You can contribute compute by AI training on the network — often called Qubic mining. QLI runs a long-established AI-training platform where you connect Windows, Linux, or HiveOS hardware and earn QUBIC for the compute you contribute.

Build on Qubic with QLI

QLI is the first Qubic service provider — 3+ years running nodes, AI-training infrastructure, and public services for the network.

Start AI Training Get in Touch