Common Patterns
Dynamic Programming
Overlapping subproblems and optimal substructure: how DP appears in AI coding problems, from edit distance to resource allocation.
Dynamic programming is the pattern that benefits most from working with AI. The model can produce a clean bottom-up implementation in seconds. Code that used to take a confident candidate twenty minutes of careful index-juggling now lands as soon as you can describe it. It can also produce DP code that's silently wrong on an empty string, or on equal strings, and the bug won't show up in the test the interviewer gave you. Your job in an AI-enabled interview is to point the model at the right recurrence and catch the bugs when it doesn't.
Recognizing the pattern
The interview prompt won't say "use DP." It'll describe a spell checker, a diff tool, a scheduling problem, or a task allocator, and you have to hear the pattern underneath. Listen for these four signals:
- The answer is "the best X," not "any X." Minimum cost, maximum profit, longest run, shortest sequence, number of ways. Optimums sit on top of DP, greedy, or both, almost without exception.
- Choices compound. The best next move depends on what you've already chosen. You're not making independent picks.
- The state has a clean description. If you can say "given the first i items and a budget of j, the best we can do is..." in one breath, you've found the state.
- Brute force is exponential. If trying every combination would be too slow at the input size given, the subproblems are overlapping and you can cache them.
Some real product wrappers and what they actually are:
- "Suggest the three closest dictionary words for a misspelled query." Edit distance: the minimum number of single-character edits (insert, delete, replace) to turn one string into another.
- "Compute a minimal diff between two config files." Longest common subsequence: the largest set of lines appearing in the same relative order in both.
- "Pack the highest-value subset of tasks into a worker's eight-hour shift." Knapsack: the best subset of items that fits within a budget.
- "Schedule non-overlapping appointments to maximize revenue." Weighted interval scheduling: knapsack but the items conflict with each other.
- "Count the valid ways to break this string into known words." Word break: count or check decompositions of a sequence.
Recognition is the place where the model can suggest but shouldn't decide. Ask "does this look like DP to you?" and the AI will tell you yes; it agrees with whatever framing you give it. The thing the interviewer is grading is you naming the pattern out loud and being right. "OK, this is edit distance, two sequences, three operations" is what shows control.
DP or greedy? Not every optimization problem is DP. If the locally optimal choice is always globally optimal, you're in greedy territory and there's a much simpler solution. Under pressure, ask yourself whether always picking the best-looking option right now could miss a better answer later. If yes, you need DP. If no, greedy works. Get this wrong and the AI will faithfully build the wrong thing on top of your wrong call.
A refresher on DP
Prompting the AI
A worked example: edit distance
Verifying the AI's code
Failure modes specific to DP
When to use vs. alternatives
What interviewers expect
Putting it together
Where to practice
Purchase Premium to Keep Reading
Unlock this article and so much more with Hello Interview Premium