Mock assessment

Mock Assessment -- Interview-Calibrated Testing

Tests calibrated to what companies actually ask. Based on analysis of 100+ take-home assignments and interview rounds.

Trigger

• /assess placement -- placement test to skip mastered topics
• /assess quick -- 15-minute concept check
• /assess milestone -- 45-minute milestone-calibrated assessment
• /assess interview -- 4-round full simulation

Assessment Modes

Placement (20-30 min per topic)

Determines if a student can skip a topic they already know. Use before starting the curriculum or when an experienced developer joins.

Process:

1. Student declares which topic(s) they want to test out of.
2. For each topic, ask 5 questions covering the full exercise range:
   • 2 concept recall (from memory, no lookup)
   • 1 design question (architect a solution using that topic's patterns)
   • 1 debugging question (here's broken code/config, what's wrong?)
   • 1 production question (how does this work at scale / in production?)
3. Score using the standard rubric.

Skip criteria:

• 8/10+: Skip the topic entirely. Record in PROGRESS.md: "Topic N: PLACED OUT (score X/10, {date})"
• 6-7/10: Skip exercises 1-5 (fundamentals), start at exercise 6+ (intermediate). Record partial placement.
• 5/10 or below: Start from exercise 1. No skip.

Rules:

• Placement is topic-level, not exercise-level. You can't skip individual exercises within a topic (except with 6-7 partial placement).
• Questions must cover the FULL topic range -- don't just test the easy exercises.
• Be ruthless. A senior Python developer who can't explain asyncio.gather vs asyncio.wait doesn't skip Topic 2.
• Record all placement results in assessments/{date}-placement-topic-{N}.md.
• If the student barely passes (exactly 8/10), warn them: "You passed, but the knowledge is thin. Consider doing the milestone project anyway."

Quick (15 min)

3 concept recall questions + 1 design question. No code.

Milestone (45 min)

Mirrors real take-home assignments. See ASSIGNMENTS.md for topic-specific assessment details.

Interview (4 rounds)

Full simulation matching real hiring:

1. Technical screen: Python fundamentals + LLM basics
2. System design: Design a RAG/agent system for a use case
3. Code review: Review provided code, find issues, suggest fixes
4. Behavioral: "Tell me about a system you built" (prior agentic AI projects)

Execution

1. Select assessment type (or ask the student).
2. Generate questions calibrated to current topic + cumulative knowledge.
3. One question at a time. Note time but don't hard-enforce.
4. Evaluate against rubric in RUBRIC.md.
5. Output results:

   ASSESSMENT -- {type} -- {date}
   
   Score: X/10
   Strengths: [specific]
   Gaps: [specific, actionable]
   Market readiness: [honest vs. job requirements]
   Next focus: [what to work on]
   

6. Save to assessments/{date}-{type}.md for tracking.

Rules

• 8/10 from this system means "would pass a real interview round."
• Don't soften feedback. "You would not pass this round" is valid.
• Connect every question to real market requirements.
• Track scores over time -- they should trend up.
• Cross-reference with spaced-review deck to find systematic gaps.

Outcome Tracking (Calibration)

Our scoring is only useful if it predicts real-world outcomes. After any real interview, take-home, CCA exam, or job offer:

1. Record the outcome in assessments/OUTCOME-TRACKING.md
2. Compare against our last assessment score for that topic
3. After 5+ entries, recalibrate if prediction accuracy drops below 80%

See OUTCOME-TRACKING.md for the full calibration protocol.