John Carmack and Rich Sutton team up to build AGI by 2030

Legendary game developer and engineer John Carmack has found his next co-op partner for artificial general intelligence.

Carmack's startup Keen Technologies announced a direct partnership with Dr. Richard Sutton, widely regarded as the founding father of reinforcement learning, to accelerate the path toward AGI.

This marks the first major public milestone for Keen since its 2022 launch with a $20 million financing round backed by tech heavyweights including Nat Friedman, Daniel Gross, Patrick Collison, Tobi Lutke, Jim Keller, and Sequoia Capital. Carmack famously stepped down from his advisory CTO role at Meta in December 2022 to dedicate his time entirely to this effort.

Betting against the LLM consensus

While the current tech world races to burn billions on GPU clusters and text scrapers, Carmack and Sutton are intentionally heading in the opposite direction.

Carmack pointed out that the AI sector is awash in capital, compute, and data, yet still dominated by fashions that might actually slow down genuine breakthroughs. Both researchers consider themselves outside the mainstream, focusing instead on agency, continuous temporal interaction, and long-lived agents that learn to predict and control signals in a complex world.

Sutton will retain his academic positions at the University of Alberta and the Alberta Machine Intelligence Institute. His ongoing work on The Alberta Plan will directly feed into the partnership.

Together, they have set an ambitious timeline: delivering a genuine AGI prototype by 2030, complete with documented signs of life for artificial cognition.

Why it matters

We think this partnership represents one of the most credible counterweights to the brute force scaling narrative dominating Silicon Valley today.

For builders and researchers, don't expect an API or consumer product anytime soon. This is pure fundamental research. But it provides serious intellectual backing to the argument that next-token prediction on frozen web corpora is a dead end for true autonomy.

Carmack is one of the greatest systems engineers alive, known for squeezing impossible performance out of constrained hardware across 3D graphics, rocketry, and virtual reality. Sutton provides the foundational reinforcement learning theory that modern AI still leans on.

Taking an agency-first, non-mainstream path carries massive technical risk, and 2030 is an aggressive target for genuine AGI. Still, pairing a legendary systems architect with the pioneer of temporal-difference learning is easily one of the most fascinating moonshots in AI right now.