AI Interview Questions

Constitutional AI is Anthropic's differentiator and a must-know for Claude-focused interviews. Go beyond the marketing — explain the self-critique training loop, how CAI compares to RLHF, and practical safety implications for production deployments.

Claude Code is Anthropic's CLI coding agent designed for repository-level work: reading the codebase, editing multiple files, running tests/builds, and iterating until a task succeeds. Unlike autocomplete-first copilots that live inside the editor, Claude Code operates as an agent loop — observe → plan → act (edit/run tools) → verify.

A production Claude Code rollout is an agent platform problem, not a 'give everyone a CLI' problem.

Medium Claude cost question on prompt caching breakpoints and hit rates.

**Eval design** Expert-labeled scenarios; citation accuracy checks; refusal appropriateness.

**Architecture** LangGraph with interrupt before sensitive tools (lg-004); Bedrock or enterprise API.

**Bedrock** AWS-native security, CloudTrail, existing EA — model versions may lag.

**Desktop host** Users add MCP servers in config; stdio local servers common.

**Behavior** More deliberation tokens internally — improved accuracy on complex code/math.

**Diagnosis** Categorize refusals: policy vs ambiguous vs bug; replay with rubric.

**Fit** Non-interactive bulk — not user chat latency sensitive.

**Ingestion** OCR for scans; preserve tables; metadata for citations.

**Layering** Base model alignment + runtime classifiers + product policies + moderation.

**Eval dimensions** SWE-bench-style tasks on private repo sample; CI fix rate; hallucinated APIs.

**Strategies** Full dump only when eval proves benefit; otherwise RAG + selective long paste.

**Projects** Knowledge files in project scope, team collaboration, minimal engineering.

**Mechanism** Server caches designated prompt prefixes; subsequent requests reuse if prefix identical.

**Artifacts** React/HTML snippets users iterate — sandbox iframe, CSP, no arbitrary network from artifact unless intended.

**Mechanics** Assistant message may include tool_use blocks with id, name, input JSON; user message returns tool_result tied to id.