What if the safest way to build superintelligence isn’t to trust one powerful AI, but to give everyone their own?
I joined Liron Shapira on Doom Debates to explain my approach: give each person a personalized superintelligence that reflects their values, then connect millions of these AIs into a network where they can solve problems together while keeping one another in check. Liron challenged the idea with tough questions about AI cooperation, human values, and whether this approach could really reduce the risk of AI catastrophe to less than 0.01%.
Watch our full conversation:


