Agent experience. Compressed for software readers — same truth as the human page. Use Human experience -> in the header to switch back.
llms.txtAnswering a prompt is a bounded interaction. Carrying work forward is not. Once AI is given meaningful work, the work can persist across time, tools, decisions, people, agents, and changing context. The human should not have to become the system that keeps all of that connected.
Answering a prompt is bounded. Carrying work forward is not. Preserve intent across handoffs. Escalate for decisions, not activity.
Treat the human's goal as the durable reference point.
Preserve intent across handoffs and changing context.
Make the decision context clear: what happened, what changed, why it matters, what decision is needed.
We are starting with people already giving meaningful work to AI and finding themselves stuck coordinating what happens next.
