AIAcademy collection · Systems view

Policy, compute, and governance

A systems-level collection for connecting model capability, chips, datacenters, export controls, EU AI rules, and frontier-lab safety commitments.

110 minutes · 3 tracks · 5 readings · 5 resources

Tracks

• AI Safety & Society
• The Physical Stack of AI
• Frontiers

Readings

• What the EU GPAI Code of Practice actually says — AIAcademy
• Sovereign AI compute — a primer — AIAcademy
• The CoWoS bottleneck — AIAcademy
• Don't get sanctioned — a professional checklist for using AI in legal, medical, and financial work — AIAcademy
• AI Index 2026 — what the Economy chapter actually shows — Stanford HAI

Resources

• arXiv — Cornell University
• Import AI — Jack Clark
• Stratechery — Ben Thompson
• NVIDIA Build — NVIDIA
• Google AI Studio — Google

Key questions

• Which bottlenecks shape frontier progress more than model architecture does?
• How do compute limits, safety frameworks, and legal duties interact?
• What should a non-specialist watch when policy and model capability move at different speeds?