Who manages the agents?

3 hours ago

The article presents two AI futures: one where AI is a deity controlled by a small clergy-like group, and another where AI amplifies human capabilities with humans at the center.
A new technical clergy is emerging—those who build frontier AI systems, have privileged access, and decide who can use them, warning of mass unemployment but seeing it as liberating.
Concerns exist about a two-tier system: powerful AI for a select few (e.g., partners, researchers) vs. constrained AI for the public, due to safety risks like bioweapons and disinformation.
In fields like software development, a small fraction of superusers achieve 100x productivity gains with AI agents, while the median sees little improvement, widening the gap.
The future should be built by billions, not just for them, emphasizing universal participation over mere access, with humans managing agents to enhance creativity and productivity.
Companies should focus on making the median worker 2x more productive through AI agents managed by employees, not on headcount reduction, to avoid disenfranchisement.
Agents should be sovereign—owned and controlled by companies, with humans at the center managing identities, memory, and skills—to ensure accountability and prevent dependency.
The article calls for rejecting prophecies of AI concentration, advocating instead for distributing AI to increase human agency and building a future where everyone becomes an agent manager.

Hasty Briefsbeta