@arcprize@AnthropicAI Naively, I figured ARC-AGI-3 would take years to be saturated. Now it feels like 6 months max
I assume ARC Prize is building for ARC-AGI-4. Meanwhile https://t.co/VgB82YUDSq is working on scaling the previous benchmarks horizontally
@fchollet Very impressive results, particularly given the modest performance by Grok 4.5 just days ago
Is there human completion/reasoning data on ARC-AGI-3 or should I add AA3 puzzles to https://t.co/VgB82YUDSq?
Old Money Music Industry Oligopoly vs. New Tech Generative AI Firms
Huge missed opportunity for benefit to both labeled artists and generative composers alike. It’s a shame tbh