A developer has released LeetCompass, an open-source visual catalog that breaks LeetCode's interview problems into granular sub-patterns, addressing what they see as gaps in existing study guides.
The tool maps problems to specific techniques like fixed vs. variable sliding windows, Floyd's cycle detection, and two-pointer variations. Users can import their solved problem history, overlay it with curated lists like NeetCode 75, and filter by target companies to assess interview readiness.
The developer's argument: existing pattern frameworks like Educative's 26 patterns or Sean Prashad's GitHub lists are useful starting points, but they lump too many problem types together. Hard problems often combine multiple patterns (graph traversal plus dynamic programming, for example), and LeetCode's own topic tags frequently miss the optimal solution approach.
This isn't the first attempt at pattern-based prep. Community estimates suggest techniques like sliding window and two pointers cover 70-80% of array and string interview questions. Multiple GitHub repositories and YouTube channels already organize problems this way, and LeetCode's own Explore section offers structured paths.
The real question is whether granular categorization helps or hurts. Pattern advocates argue it builds recognition speed, turning what feels like 3,000 random puzzles into maybe 50 templates. Skeptics counter that over-relying on patterns can backfire when problems don't fit neat buckets, and that grinding through LeetCode's editorial solutions builds deeper problem-solving instincts.
The tool is live at leetcompass.io with the taxonomy still incomplete. The developer is taking pull requests for new patterns and reclassifications.
What this means in practice: If you're prepping for FAANG-style interviews, this is another lens to try. The visual format might help with pattern recognition. But it's solving a second-order problem (how to organize practice) rather than the first-order one (whether LeetCode-style interviews measure what companies actually need). The taxonomy will live or die based on whether people find it clearer than alternatives.