Google's famous 20% time policy let engineers explore passion projects one day weekly. That culture-driven approach worked because it was protected, unmeasured time. Today, AI is resurrecting the same exploratory spirit, but without dedicated hours. Developers now chase side projects during gaps between agent tasks, creating what the author calls 120% time - constant work plus exploration.

The constraint has flipped from hours to attention. Coding once felt like metalwork, but AI agents make it clay-like, reshaping outputs cheaply. This frees mental space for speculation, yet demands the effortful kind of attention - holding context across multiple agent threads. Harvard Business Review labels this AI brain fry: supervisor fatigue from systems moving faster than human cognition can track.

The optimism isn't universal. While developers gain exploration room, other fields face displacement. Illustrators and writers report AI undercutting rates and replacing work trained on their content. The same technology that buys room for some takes it from others. Attention becomes the finite resource nobody can allocate freely.

The original 20% time was a luxury paid for by Google's profitability. Its replacement exists under constant measurement. Fifty years of productivity gains have flowed upward, not to workers. The question isn't whether AI saves labor - it does for those steering the tools. The question is who collects that dividend.