Making AI agent work survive crashes, restarts and session limits, so progress can be resumed without repeating or losing work.
One small task, followed through a crash, a corrected input and a retried side effect: which records let someone else carry on without losing progress, repeating what cannot be undone, letting a presumed-dead worker write, or trusting a result built from the wrong input.
Graph engineering and orchestration: what an arrow means, which work can overlap, why a ready task still waits, what survives an interruption, and who decides a finished result can be used.
Long-horizon AI work fails when the session is the system. Decisions, drafts and disputed sources need a production form that outlives the chat.
We run enough automated, long-running work that we built layers whose whole job is to tell us what state everything is in. That was supposed to solve the problem. It didn't.
A durable visual-production workflow for traceable jobs, retained results and independent review.
A recoverable orchestration system for multi-step production work, using dependency state, material evidence and human review to decide what can proceed.
A recoverable 3D reconstruction workspace that turns image or video references into inspectable geometry for further modelling or printing.
Still-to-motion work is accepted as a durable media job, so generation can finish independently of the request that started it and the result can be retrieved later.
A persistent messaging layer for durable, searchable handoffs between independent working sessions.
A historical source-to-draft workflow that kept research, drafting, section-aware media and editorial decisions attached to durable workflow state.