ARC-AGI-4 announced: a benchmark for open-ended innovation where humans still crush AI
GregKamradt · x · 2026-09-13
ARC Prize announced ARC-AGI-4, a benchmark for autonomous open-ended innovation, continuing its commitment to open source and giving the research community a shared target for progress. The org's argument: despite rapid model progress, humans still significantly outperform AI at open-ended invention — the meta-skill that unlocks progress across every field, and the foundation for a positive-sum future.
Greg Kamradt adds two points: (1) open source maximizes collective progress and safety by spreading knowledge of how intelligence is built, not just the tech itself; (2) open innovation is the last frontier where humans still beat AI, and it needs transparent tools to measure — which is exactly what ARC-AGI-4 targets.
More from AGI Musings
- Thom Wolf agrees with 75% of Amodei's letter, but calls its lead-widening framing counterproductive — beffjezos · 2026-09-13
- e/acc founder: AI capability centralization is the opposite of safety — beffjezos · 2026-09-13
- Armin Ronacher pushes back on frontier-pausing in new P(doom) essay — mitsuhiko · 2026-09-13
- AI math proofs won't kill understanding: post hoc exploration keeps mathematicians central — njyx · 2026-09-13
- Andreessen: rogue AI cyber attacks smell like false flags, a convenient excuse for regulatory capture — beffjezos · 2026-09-13
- Sam Altman: stopping AI would mean kids dying of curable diseases — firstadopter · 2026-09-13