ARC Prize President Maps Post-AGI World

Kamradt’s framing is compelling because it treats AGI as an optimization and deployment problem, not a single product launch. The boldest claim is that the algorithmic advantage will evaporate quickly, leaving infrastructure and data access as the real battleground.

• –Stateful AGI would make continual learning and persistent context core developer primitives.
• –A low-compute baseline paired with optional extra inference suggests a future where capability scales with budget, not just model size.
• –His robotics timeline is more cautious than typical AGI forecasts: software intelligence must mature before general-purpose physical labor becomes economical.
• –The predicted physics-discovery wave highlights a near-term research opportunity for agents that can synthesize existing literature and evidence.
• –The thesis also reinforces ARC Prize’s focus on adaptive intelligence rather than static benchmark performance.