The 21 Topics AI Can't Replace (and How to Find Your Own Gaps)

AI can write thousands of lines of code in minutes. Then it tells you everything works fine.

Who checks that it's true? You do.

That is the part AI can't replace. It's not typing speed. It's judgment: knowing what to build, why it can fail, and what it will cost you in three years.

Below are 21 topics that build that judgment. For each one, there is one question. Try to answer it out loud, without Google. If you can't, you found a gap.

1. Design foundations

Software Architecture. When do you choose a monolith, and when microservices? What will each choice cost in 3 years?

Design Patterns and SOLID. How do you add a feature without breaking ten old ones?

Domain-Driven Design. Why can code that matches the ticket still solve the wrong problem?

Requirements Engineering. What question is the client not asking, but should be?

Quality Attributes and Trade-offs. Can you name a decision where you traded cost for performance?

2. Core computer science

Data Structures and Algorithms. Why can an algorithm that is fast on 1,000 records fail on 100 million?

Concurrency and Parallelism. Two pieces of correct code can still corrupt shared data. Why?

3. Data

Database Design. How does an index speed up reads but slow down writes?

Caching Strategies. What is a cache stampede, and can a wrong cache be worse than no cache?

4. Systems at scale

System Design. What is the difference between "it works" and "it stays working under pressure"?

Distributed Systems. Why does code work on one machine and fail on many servers?

Reliability Engineering. What is idempotency, and how does it stop a double payment?

Performance and Capacity Planning. How do you know if the bottleneck is the code or the database?

5. Delivery and operations

Testing Strategies. Does a passing test prove the system behaves correctly?

DevOps, CI/CD and Releases. How do you roll back a database migration?

Observability and Monitoring. Why can health checks stay green while the system is failing?

Disaster Recovery. Have you ever tested restoring a backup?

6. Security and mindset

Security Engineering. How can one small permission bug expose millions of records?

Technical Debt and Maintenance. Which quick decision today becomes debt tomorrow?

Problem Solving. Can you tell a symptom from its root cause?

Software Engineering Ethics. What are you responsible for when your system holds people's money or data?

How to use this list

1. Answer each question honestly.
2. Rate yourself from 1 to 5. (1 = never heard of it. 3 = I can explain it. 5 = I can teach it and defend my choices.)
3. Pick your 3 lowest topics and study them for 30 days.
4. Build one small demo for each. Add a circuit breaker. Add an index and measure it. Break it on purpose.
5. Explain your trade-offs out loud, then re-score yourself.

A few free places to start: Refactoring.Guru for patterns, Use The Index, Luke for databases, the System Design Primer for scale, and the Google SRE books for reliability.

Get the full kit

I turned this list into a 26-page Self-Assessment Kit. Every topic has 3 honest questions, the best free study links, and a rating row. It also has a scorecard (out of 105) and a 90-day plan.

👉 Get the kit on Gumroad

Which topic is your weakest? Tell me in the comments.