Failure Modes &
Usability Patterns
Identifying the friction points between human cognition and artificial intelligence.
A library of common failures to avoid when building accessible agents.
Over-interruption
When AI assistants break meaningful flow instead of preserving it.
Cognitive Overload
The friction caused by AI providing too much context or too many options.
Silent Failure
When systems hallucinate or fail without clear error signals, eroding trust over time.
Tool Sprawl
The fragmentation fatigue of managing disconnected, specialized agents.
Stateless Amnesia
Agents that forget everything between sessions, forcing users to re-establish context from scratch every time.
Prompt-Only Governance
Treating safety as an instruction to the model rather than a structural guarantee. Fails the moment the model misbehaves.
Unverifiable Autonomy
AI agents that act in consequential ways without leaving a tamper-evident audit trail. No proof of what happened.
Credential Trust Collapse
Multi-agent systems where trust is granted by credential alone. A stolen token is indistinguishable from a legitimate one.