
Agent Co-workers and Multiplayer Work

AI’s third era: the rise of persistent AI coworkers | Tara Seshan (OpenAI’s product lead)
Topic 4 of 17
Published Aug 30, 2026 · Clip 15:36–20:06 (4:30)
Tara describes two shifts: working with agents at higher levels of abstraction — 'let the agent cook' — and treating persistent agents as co-workers you sync with like teammates. She also sees a move from 1:1 human-agent work toward collaborating with agents together as a team, since sharing screenshots of Codex threads on Slack wasn't natural collaboration; ideally work becomes a multiplayer game of steering agents. When Lenny frames this as a slow progression of trust and a slow takeoff, Tara adds that beyond intelligence and long-task stamina, tactical 'meat and potatoes' enable agents: local machine access, cloud infrastructure, and access to third-party systems — an agent locked away from docs, Slack, and databases is useless.