Rethinking Technical Interviews - What Blackjack-Interview Taught Me About Hiring in the AI Era

My view on interviews changed when I built blackjack-interview. It showed me what hiring should measure in today's world.

I used to watch companies scramble to block AI in interviews. Treat it like a cheat code. But that misses the point entirely. The risk isn't tools. It's asking the wrong questions.

Flipping the Frame

With blackjack-interview, I flipped the question. Instead of "can you write this by hand without AI?", I asked: "can you think with AI and still deliver value?"

The challenge gives candidates:

• A skeleton game implementation
• 40 failing tests
• TypeScript types
• A one-hour timeframe

Then it says: go. Use whatever tools you want. Your job: ship something correct, readable, maintainable.

What I'm Actually Testing

I'm not testing memorized algorithms or syntax speed. I'm testing:

• Judgement - Can they verify AI suggestions against actual requirements?
• Prompt-crafting - Can they get useful output from AI tools?
• Debugging ability - Can they fix AI output when it's wrong?
• Shipping under real conditions - Can they deliver working software with modern tools?

The codebase intentionally contains misleading comments. Candidates who blindly follow AI suggestions without testing will fail. Those who verify suggestions against test requirements will catch the errors.

Why This Matters Now More Than Ever

Technical interviews historically focused on toy problems: algorithms, puzzles, whiteboards. These rarely reflect real work. AI didn't break interviews. It exposed how off-base they already were.

If you build the challenge around realistic context, real constraints, and tools that mirror the actual job, you get better signals:

• Candidates who integrate with modern workflows
• Early visibility into code structure, testing habits, and edge-case handling
• Less bias toward pattern memorizers

The Evaluation Framework

The interview guide breaks evaluation into five areas:

AI Tool Proficiency (25 points)

• Excellent: Verifies suggestions, catches misleading comments, iterates thoughtfully
• Poor: Blindly follows AI suggestions, struggles to debug output

Test-Driven Development (20 points)

• Excellent: Constant test running, reads tests before coding, systematic progression
• Poor: Rarely runs tests, implements without verification

Code Quality (20 points)

• Excellent: Clean, readable, handles edge cases, proper TypeScript usage
• Poor: Messy code, significant issues

Problem-Solving (20 points)

• Excellent: Breaks down problems, systematic debugging, independent solutions
• Poor: Cannot decompose problems, struggles without hints

Speed & Efficiency (15 points)

• Excellent: Completes in under 45 minutes, all tests passing
• Adequate: 60+ minutes, partial completion

Success Benchmarks

Minimum Bar (Pass): 70%+ tests passing, core game works, demonstrates AI usage, reasonably clean code

Strong Performance: 90%+ tests, full functionality, correct Ace handling, effective AI tool use, under 60 minutes

Exceptional: All tests passing, polished implementation, sophisticated AI collaboration, excellent code quality

Red Flags to Watch For

• Blindly following misleading comments without testing
• Copy-pasting AI code without understanding
• Never running tests or reading requirements
• Cannot debug failures independently

These behaviors predict how someone will work day-to-day. Better to see them in an hour than discover them after three months.

What This Doesn't Cover

Blackjack-interview isn't perfect. It doesn't cover:

• System scalability
• Large-team coordination
• Architectural decisions at scale

But as a live test, it gives you a window into how a candidate works today, with today's tools.

The Lesson for Hiring Managers

Change the interview question. Measure signals that matter.

• Judgement over memorization
• Adaptability over rigid process
• Code quality under modern constraints

Because in 2025, tools have changed. But the need for people who can deliver real work hasn't.

---

Try it yourself: The blackjack-interview challenge is open source. Use it for your next hire, or take it as a candidate to see where you stand.