If you're still optimizing for individual capability — human or model — you're solving last decade's problem. The real multiplier is team architecture, and agents just reshuffled every variable in that equation.
In his essay, Avichal Garg makes the case that obsessing over 10x individuals is the wrong frame entirely. What compounds is the team: the trust between people, the way their strengths interlock without redundancy, and the friction cost of coordination kept ruthlessly low. A brilliant hire who generates drag — through miscommunication, misaligned incentives, or the simple overhead of managing their ego — can quietly destroy more value than a merely good hire who fits cleanly into how the group thinks and moves.
Agents collapse certain coordination costs Garg worried about — they don't need status meetings, don't protect turf, don't go silent because they're offended. But they introduce new ones: handoff failures between agents, trust gaps between human decision-makers who can't read an agent's reasoning the way they'd read a colleague's face, and the compounding risk of misaligned defaults nobody caught early. The 10x team principle holds and actually sharpens. You still need complementary strengths, clear ownership, and low drag — you just have to design those properties deliberately into a system that won't self-correct socially the way humans eventually do.
- Hire for team fit more than individual ceiling, then apply that same logic to how you compose agent workflows
- coordination cost doesn't disappear with agents, it just moves somewhere less visible
- trust is an architecture decision now, not a culture outcome.
