Day 5 — DSA When Everyone Uses Copilot
GitHub Copilot can solve most LeetCode Easy problems and many Medium ones. ChatGPT can solve Hard problems with enough context. Every interviewer knows this.
So why are companies still testing DSA?
Because they are not testing whether you can solve the problem. They are testing whether you can think through a problem in front of another person, in real time, without a computer doing the thinking for you.
What Actually Changed
What changed: AI can generate the code. So the code itself is table stakes.
What did not change: Your ability to reason about the problem before writing any code.
Interviewers at every company — from TCS to Google — have shifted toward asking more questions during the problem-solving process.
Before: "Here is the problem. Write the code."
Now: "Here is the problem. Walk me through your thinking before you write anything."
The student who immediately starts coding looks like they are reciting a memorised solution. The student who says "Let me think through this — a brute force approach would be O(n²) but I think we can improve it using a hashmap..." looks like an actual engineer.
What Interviewers Are Looking For Now
For service companies (TCS NQT, Infosys, Wipro): The format has not changed much. Still MCQs and short coding questions in an automated test. What changed is the difficulty calibration — they have made tests slightly harder because they know AI can solve easy problems.
Focus: Data structures basics, time complexity, logical reasoning. Understand these concepts genuinely. The tests are auto-graded but some companies do manual review of borderline candidates.
For mid-tier product companies (Zoho, Freshworks, Persistent): Live coding rounds with an interviewer watching. They ask you to explain your approach before and during coding. They ask follow-up questions: "What if the input array is sorted?" "What if we need to handle duplicates?"
These questions test adaptability. AI cannot answer them in real time during an interview.
For top companies (Amazon, Microsoft, Google): Heavy focus on explanation. The interviewer often does not care if the first solution is optimal — they care that you can identify that it is not optimal and iterate toward better.
How to Actually Prepare
Step 1 — Learn the patterns, not the solutions
There are roughly 15 core DSA patterns that cover 80% of interview questions:
- Two pointers
- Sliding window
- Fast and slow pointers
- Merge intervals
- Cyclic sort
- In-place reversal of linked list
- Tree BFS
- Tree DFS
- Two heaps
- Subsets
- Modified binary search
- Top K elements
- K-way merge
- Dynamic programming
- Backtracking
If you understand when and why to apply each pattern, you can approach unseen problems. Memorising solutions cannot do this.
Step 2 — Practice explaining before coding
For every LeetCode problem you attempt:
- Read the problem
- Speak your approach out loud (or type it out) before writing any code
- State the time and space complexity of your approach before coding
- Then code
This trains the skill that interviewers actually test: articulating your reasoning.
Step 3 — Do timed practice without AI
Once a day, pick one problem you have not seen before. Set a timer for 20 minutes. No AI. No hints. Just you and the problem.
If you cannot solve it in 20 minutes, look at the approach (not the solution), try again for 10 more minutes, then study the solution.
The daily timer practice is the most important habit. Without it, you will not know what you actually understand versus what you only recognise.
How to Use AI Correctly for DSA Prep
Use AI to understand patterns after you have struggled:
I tried solving the sliding window maximum problem and
I got a solution that works but it is O(n²).
I know the optimal solution uses a deque but I do not
understand why. Can you explain the intuition behind
using a deque here, without showing me the full code first?
This is productive. You struggled, you have a partial understanding, you are asking for the missing insight.
Do not use AI to generate solutions before attempting:
If you look at solutions before genuinely attempting the problem, you are training your pattern recognition — not your problem-solving ability. You need both. But pattern recognition without problem-solving fails in live interviews.
The Complexity Question You Will Always Get
Every technical interviewer asks: "What is the time and space complexity of your solution?"
Many students memorise answers without understanding them. This fails on follow-up questions like "Why is it O(n log n) and not O(n²)?" or "How would the complexity change if the input was already sorted?"
Practice this sentence structure for every problem you solve:
"This solution is O([time complexity]) because [specific reason]. The space complexity is O([space complexity]) because [specific reason]."
The "because" is what matters. Any student can say O(n log n). Fewer can explain why.
For Students at Tier-2 and Tier-3 Colleges
If you are at a college that does not have TCS or Infosys visiting for on-campus placements, the DSA bar for off-campus applications is actually manageable.
For off-campus applications to service companies: LeetCode Easy level comfort is enough. 50 problems across the main patterns, understood well, is more than sufficient.
For off-campus applications to mid-tier product companies: LeetCode Medium level. 30-40 problems, focused on the most common patterns — arrays, strings, trees, graphs.
The students who do not get placed are usually not the ones who could not solve Hard LeetCode problems. They are the ones who could not solve Easy problems under pressure because they never practiced with a timer.
Your Action Item
Pick one pattern from the list of 15 above that you feel least confident about. Find three LeetCode Easy problems that use that pattern. Solve them today without AI assistance, explaining your approach out loud before coding.
That is the entire action. One pattern. Three problems. Spoken explanation before each.
Day 5 of 15 — AI Survival Kit for Engineers