Why can I only learn which coordination failure hit my agent pipeline after it already did?
Opportunity
The MAST taxonomy, validated on over 1,600 production agent execution traces at NeurIPS 2025, names 14 coordination failure modes that account for 79% of multi-agent system failures. Knowing the taxonomy after a failure is useful. Knowing which failure mode a specific architecture will hit before deployment does not yet exist as a tool or discipline. Current evaluation practice runs component-level tests on individual agents but has no method to characterize the emergent failure profile of the whole system under realistic coordination load. Practitioners face a consistent 37% gap between lab performance and production reliability with no diagnostic that predicts where it will open.
Why it matters
Predictive coordination failure profiling is what turns a post-mortem taxonomy into a pre-deployment design check, which is the actual lever for cutting the production gap.
How I score the opportunity
The Opportunity Score is my own read, not a measurement: how much it hurts, how often it bites, and how little exists to solve it today. Higher means I think it is more worth building.
How much pain it causes when it shows up.
How often people actually run into it.
How little good tooling exists for it today.
More problems worth solving
Why does every AI app forget me the moment I close the tab?
AIWhy is learning a new field still gated by knowing what to ask?
AIWhy can a non-expert not verify what an AI just told them?
AIWhy do we test models on benchmarks but ship them on vibes?
AIWhy do AI agents have no memory of their own mistakes?
AIWhy can't I audit what a model was actually trained on?