Patterns

Companies don't pick AI coding problems at random. The reason you see the same kinds of problems over and over is deliberate: they're challenging enough to allow for multiple paths and non-obvious solutions, but not so complicated they can't fit in a 45-minute interview. You're less likely to see simple LeetCode easy/medium problems, because AI squashes those immediately and there's nothing interesting left to evaluate.

How patterns change in an AI interview

In a traditional interview, knowing patterns saved you time. In an AI-enabled interview, knowing patterns is what lets you steer the AI toward the right approach and follow the solution it hands back — instead of nodding along to code you don't fully understand. The AI handles the implementation. What you bring is the judgment behind it. Think of it as more bandwidth for the conversation rather than ready-made templates.

This is a double-edged sword. One mistake we see constantly: candidates have a vague, wrong idea of how to solve a problem, ask the AI with some specificity, and get a contorted version skewed to what they asked for. If you ask for a "sliding window" solution to a graph problem, the AI will work really hard to satisfy your request even when it's not the right approach.

The difference between candidates who ace these interviews and those who struggle is how quickly they can:

1. Map the problem to a known approach.
2. Communicate that approach to the AI with enough specificity that the output is useful.
3. Understand and verify outputs that might be genuinely deep.

You won't learn Monte Carlo tree search during the interview. But if you know it in advance, you're far more effective when it shows up — either in the AI's response or as the obvious solution to the problem.

The patterns

Here's a quick tour. Each one gets a full deep-dive in the Common Patterns section.

Pattern	When it shows up
Graph Search & Pathfinding	Navigating networks, shortest paths, exploring connected structures. BFS, DFS, Dijkstra, A*.
Topological Sort	Anything with dependencies: build systems, course prerequisites, task scheduling. If things must happen in order, you're looking at a DAG.
Backtracking	Constraint satisfaction. Configuration generators, puzzle solvers, scheduling with complex rules. Try, check constraints, undo.
Greedy & Bin Packing	Where local optimization gives a good-enough global answer. Resource allocation, interval scheduling, packing.
Dynamic Programming	Overlapping subproblems. Edit distance, constrained allocation, path counting. "Have I solved a smaller version of this before?"
String Matching & Parsing	Pattern matching, text processing, expression evaluation, template engines.
Data Structure Design	Composing primitives — maps, heaps, queues — into custom structures under performance constraints. LRU caches, time-based key-value stores.
API Contracts & Side Effects	A multi-file service that's already broken. No algorithm to find — the work is matching an API spec exactly and spotting the downstream work a state change forgot to trigger.

Many problems pull in more than one pattern. Recognizing the shape is your job — the AI is there to help you build it.

The last row is the odd one out, and it's growing fastest. Instead of a blank editor you get a running service with a failing test suite, and the question becomes why is this behaving wrong rather than what algorithm applies. It rewards reading over writing, which is exactly the skill AI assistance makes scarce.