Treating agent state machines as durable execution workflows is the only way to scale real workloads. Without checkpointed intermediate states and deterministic replay, non-idempotent side effects like external mutations will continually break production systems under transient API failures. Well framed.
Would love to see a use case study comparing the implementations of 3 solutions together. This would give us even better idea of where Inngest shines.
noted, thanks for the feedback
Treating agent state machines as durable execution workflows is the only way to scale real workloads. Without checkpointed intermediate states and deterministic replay, non-idempotent side effects like external mutations will continually break production systems under transient API failures. Well framed.