Everything on Scaling Dynamics

The team identified training compute as the primary bottleneck for scaling Atlas, rather than data or architecture. Performance improves significantly with larger models and longer training times.

Impact: Indicates a clear path for future performance gains through increased investment in GPU infrastructure, suggesting that the current model is far from its theoretical ceiling.

— from WorldLab Atlas: Spatial Intelligence and New View Prediction · a16z Podcast· Sep 04, 2026