Aigarth launches distributed AI training fleet

Qubic is attempting a radical departure from the GPU-heavy status quo by betting on ternary logic and evolutionary growth.

• –Dual workload model proves Scrypt ASICs and CPU/GPU AI training can coexist at scale without resource contention
• –Shift to ternary logic (+1, 0, -1) aims for biological brain-like efficiency, potentially lowering the massive energy cost of binary AI
• –Academic validation via IEEE AMLDS 2026 acceptance lends needed scientific credibility to the project's non-standard ML approach
• –Release of the 1.12 TB Neuraxon2LifeTS dataset signals a transition from theoretical research to large-scale production training
• –Major hurdle remains proving evolutionary training can actually match gradient descent for complex transformer-scale tasks