Compounding errors
The way an agent's early mistake becomes the premise of its later actions, because the agent feeds its own output back in as the next step's input, so failures grow into runaway sequences of individually plausible steps rather than resetting like an assistant's independent answers.
Defined in 2 GAGE programs, which carry 3 distinct definitions of it. The wording above is taught in AI Governance: Applied Mastery.
How each discipline defines it
The same term does different work depending on who is using it. These are the definitions as each program teaches them, unedited.
The way an agent's early mistake becomes the premise of its later actions, because the agent feeds its own output back in as the next step's input, so failures grow into runaway sequences of individually plausible steps rather than resetting like an assistant's independent answers.
The property of multi-step agent work whereby per-step error rates multiply across steps, so a high per-step reliability becomes low end-to-end reliability (about 90 percent per step is roughly 35 percent over ten steps), which is why stops matter.
The tendency of an agent's mistakes to cascade, because a small error early in a chain of self-chosen steps can corrupt every step that follows. A core reason agent reliability falls as tasks get longer (Anthropic, 2024).
Where it is taught
The exact lessons this term appears in. The first 7 topics of every program are free with a free account.
- AI Agents vs. AI Assistants: Understanding Autonomy · Agentic AI and Workforce Integration, AI Literacy & Professional Conduct
- Communicating With and About AI Agents in Your Workflow · Agentic AI and Workforce Integration, AI Literacy & Professional Conduct
- What changes when the AI acts instead of answers: the agent risk model · Agents Under Command, AI Governance: Applied Mastery
Terms it appears with
Not an alphabetical neighbourhood: these are the terms taught in the same lessons, ranked by how often they appear together.