Claude Certified Associate — Foundations glossary
10 terms the exam expects you to know, defined in plain language.
D
- Detailed error messages
- Error messages that contain specific information enabling automated recovery decisions.
- Distractors in Solution Design
- Plausible but overengineered or misapplied solutions that fail to address the root cause of a problem.
F
- Failure Categories Distinction(distinguishing failure categories)
- The practice of differentiating between distinct failure types (e.g., timeout vs empty result, syntax vs semantic errors, drift vs overflow).
I
- Infrastructure-Level Fixes(infrastructure-level fixes)
- Additions of caching, batching, or larger context windows to address performance or context-size issues.
M
- Model-Level Fixes(model-level fixes)
- Use of separate models (e.g., classifiers, vector stores) to handle specific sub-tasks or improve accuracy.
P
- Parallel subagents(parallel subagents, parallel subagent execution)
- Subagents that execute tasks concurrently from multiple Task calls in a single turn.
- Programmatic Enforcement
- Code-level safeguards used in high-stakes scenarios to ensure correct behavior, as opposed to prompt-level nudges.
- Prompt-Level Fixes(prompt-level fixes)
- Adjustments to prompts, descriptions, or examples to influence model behavior without adding infrastructure.
R
- Right-Sized Solutions(right-sized solutions)
- The principle of selecting the simplest adequate fix (e.g., description rewrite) over overengineered infrastructure (e.g., separate classifiers).
S
- Specific error types for recovery(specific error results)
- A category of errors that convey enough detail to enable appropriate recovery actions, as opposed to generic failure messages.