A runtime that composes task execution graphs across heterogeneous AI agents. It routes each subtask to whatever agent is currently best at it. Sessions never share context. An output ships only after other agents fail to break it.
Six layers, each answering one constraint. Registry, planner, scheduler, context, validation, synthesis. Written in Lispex.
Anonymized records from internal runs, and a hash-chained session dossier you can verify yourself.
Coordination, not capability, is the bottleneck at production latency. Four lines of work.
Lena Code and Topaz. The commercial engine and the closed language the runtime was shaped by.