Blind 75 or Coding Patterns: Which Should You Study First?

Study the patterns first. Use Blind 75 afterwards, as a test of whether the patterns took hold.

The reason is simple. Blind 75 is a list of 75 specific problems. Coding patterns are the reusable techniques that solve those problems and several hundred others. If you learn the techniques, the list becomes a way to check your work. If you learn the list first, you finish it able to solve those 75 problems and no others, which is not what the interview asks of you.

This article explains the difference, shows what each approach produces, and gives an order that works.

What Blind 75 is

Blind 75 is a curated list of 75 LeetCode problems posted in 2018 by Yangshun Tay, then an engineer at Facebook, on the anonymous forum Blind. The goal was to give people a bounded, high-value set of problems instead of an unbounded grind through thousands.

It worked, and the list became the default answer to "what should I solve". It covers arrays, strings, linked lists, trees, graphs, intervals, binary search, heaps, backtracking, and dynamic programming.

Two facts about the list are worth knowing and are usually left out.

First, the author stopped recommending it. In 2022 Yangshun published a successor, Grind 75, and described it as a better version of the original. The Blind 75 was written for a specific moment and was never revised.

Second, it was designed as a shortlist for people who already knew how to solve problems. It was a way to cut down practice volume, not a way to learn technique from zero. Most of the people using it now are using it for the opposite purpose.

What coding patterns are

A coding pattern is a reusable approach that applies to a family of problems that look different on the surface.

Sliding window is a pattern. Once you know it, every problem asking for the best contiguous subarray or substring under a constraint becomes the same problem with different labels. Two pointers is a pattern. Fast and slow pointers, merge intervals, cyclic sort, tree breadth-first search, tree depth-first search, topological sort, two heaps, subsets, modified binary search, top-k elements, k-way merge, union find, backtracking, and the dynamic programming families are all patterns.

The full set runs to about 42 patterns, of which roughly 32 cover the common cases and 10 are advanced. That number is the important part. There are not thousands of interview problems. There are a few dozen techniques, wearing thousands of costumes.

The difference that decides the order

Here is the difference stated plainly:

A problem list teaches recognition. A pattern teaches transfer.

Recognition means you have seen this exact problem before and remember what you did. Transfer means you have never seen this problem and can still identify which technique applies.

Interviews test transfer. The interviewer is not going to hand you problem 42 from your list. They will hand you something adjacent, or something with a twist, and watch what you do in the first three minutes. If your preparation was recognition, those three minutes are silent.

This is why the "I solved 300 problems and still failed" story is common. Volume without technique builds recognition and nothing else, and recognition does not survive contact with an unfamiliar question.

What each order actually produces

If you do Blind 75 first, with no pattern training:

You solve the first ten problems by reading the solution, because you do not yet have the tools. You solve the next thirty with partial recall of what you read. You finish the list in six weeks. You feel prepared. Then an interviewer asks a problem that is a sliding window with a small variation, and you do not see it, because you learned "the answer to Longest Substring Without Repeating Characters", not "sliding window".

If you do patterns first:

You spend three or four weeks learning techniques, working two or three problems per pattern, with the goal of being able to state when the pattern applies and why. This feels slower. Then you open Blind 75 and find that most of it is already solvable, because you are no longer meeting 75 new problems. You are meeting 75 instances of things you know.

The second path is faster in total time, and it is the only one that produces transfer.

The order that works

Phase 1. Learn the patterns, two to three problems each.

For each pattern, you want to be able to answer three questions without looking anything up: what does this pattern do, what does a problem look like when this pattern applies, and what is the time and space cost. Solve two or three problems per pattern, not twenty. The goal at this stage is coverage, not depth.

This is what Grokking the Coding Interview is built to do: it organizes problems by the technique that solves them rather than by data structure or difficulty, which is the ordering that makes the transfer happen.

Phase 2. Use Blind 75 as an exam.

Now open the list. For each problem, do not start coding. Spend sixty seconds naming the pattern first. Then solve it.

Score yourself on the naming, not the solving. If you named the right pattern and then got the implementation wrong, that is a small gap and you close it with practice. If you could not name the pattern, that is a real gap and you go back to Phase 1 for that pattern specifically.

We keep a worked version of the list at Blind 75 with solutions.

Phase 3. Add volume where you are weak, under time pressure.

Only now does volume help, because now it is reinforcing a technique rather than substituting for one. Solve under a 30-minute clock, out loud, explaining as you go. The talking is part of what is graded.

What Blind 75 covers and what it misses

The list is good coverage, not complete coverage. Mapped against the pattern set, it is strong on some families and thin on others.

If Blind 75 is your only preparation, the gaps in the lower half of that table are the ones that show up in an interview and surprise you. Patterns cover them; the list does not.

Blind 75, Grind 75, NeetCode 150, LeetCode 75

These lists get compared constantly, and the comparison matters less than people think, because they are all the same kind of object.

• Blind 75 is the original, 75 problems, unmaintained.
• Grind 75 is the author's own successor, larger, with a schedule generator that adapts to how many weeks you have.
• NeetCode 150 is 75 plus another 75, organized by category, with video solutions.
• LeetCode 75 is LeetCode's own study plan on their site.

Any of them works as a Phase 2 exam. None of them replaces Phase 1. Picking between the lists is a small decision people spend a lot of time on; picking between "list first" and "patterns first" is the large decision, and it is the one that changes outcomes.

How long each phase takes

For someone with working knowledge of data structures who is preparing part time:

If you have less time than that, compress Phase 1 rather than skipping it. Twenty patterns understood beats seventy-five problems memorized.

Frequently asked questions

Is Blind 75 still relevant?

Yes, as a problem set. The problems are still asked and the coverage is still reasonable. What has changed is that it is unmaintained and its own author moved on to a successor. Treat it as a solid checklist, not as a curriculum.

Can I skip patterns if I have a week?

With one week, do a compressed pattern pass on the top ten or twelve patterns and then solve selectively from Blind 75. Do not attempt all 75 problems in a week; you will produce recall that fades in days.

How many problems do I need to solve in total?

Between 100 and 150 solved with the pattern named first is more effective than 400 solved by recognition. The count is not the metric. The metric is whether you can name the technique on an unfamiliar problem.

Blind 75 or NeetCode 150?

Either. NeetCode 150 has broader coverage and video explanations; Blind 75 is shorter. Both are Phase 2 material.

Should I memorize solutions?

No. Memorize the pattern trigger, which is the sentence that tells you when the technique applies. "Contiguous subarray with a constraint" triggers sliding window. That sentence is what transfers; the code is not.

Related reading

• Top LeetCode Patterns for Coding Interviews
• Don't Just Grind LeetCode
• Is LeetCode Premium Worth It for System Design Prep?