Your calendar picks the list, not the list's reputation. All three cover the same core families and any of them works if you finish it, so the only real question is whether you can. Under ten days, start none of them: run the 30-problem subset named below. Two to six weeks, take Grind 75 with the generator set to your actual weeks and hours so it trims the plan for you. Eight weeks or more, take NeetCode 150, because the extra 75 problems buy a second and third rep inside each pattern, which is what the longer list is really selling.
The decision table
Read the row for your deadline, then the column for where you honestly are today. "Solved" means unaided, under a clock, within the last year. Problems read about do not count.
| Time until the round | Rusty, or under 30 solved | 50 to 150 solved already | Bolt on |
|---|---|---|---|
| Under 10 days | The 30 below, in order, three a day, timer visible. Accept that four or five will beat you, and open the editorial at minute 30 rather than at minute 90. | The same 30, but only the ones you cannot re-derive inside four minutes. Most people here clear it in five sessions. | One 45-minute mock where you narrate out loud. At this deadline, silence fails more people than missing coverage does. |
| 2 to 6 weeks | Grind 75, generator set to your real weeks and real hours. It orders by difficulty progression, so week 1 is winnable and you do not quit on day 3. | Grind 75 at a shorter week setting than you think you need, then spend the days you recover on timed mocks. | Intervals and monotonic stack, which both 75-length lists treat as a footnote. |
| 8 weeks or more | NeetCode 150. The roughly-three-per-pattern rhythm is what turns a technique you have read into one you can retrieve cold. | NeetCode 150, skipping anything you solve inside ten minutes, and adding your target company's tagged questions for the last fortnight. | System design or SQL if your loop has one, plus the behavioural round, which nobody schedules and everybody regrets. |
| No date yet | NeetCode 150 at two a day. With no deadline to optimise against, take the list with the most repetition in it. | Company-tagged questions and mocks, not another curated list. A generic list is no longer your binding constraint. | One weekly session on the family you avoid. Everyone has one, and interviewers do not honour your preferences. |
The row matters more than the column. Someone rusty with eight weeks is in better shape than someone sharp with five days, because time converts into coverage and confidence converts into nothing.
The 30, for anyone inside ten days
Ordered so the easiest representative of each family comes first. Cap every problem at 30 minutes and read the editorial the moment the cap hits. One clean rep per family is the goal, not a perfect score.
- Hashing: Two Sum, Group Anagrams
- Two pointers: Valid Palindrome, 3Sum
- Sliding window: Longest Substring Without Repeating Characters, Minimum Size Subarray Sum
- Array transforms: Product of Array Except Self, Maximum Subarray
- Binary search: Search in Rotated Sorted Array, Koko Eating Bananas
- Stack: Valid Parentheses, Daily Temperatures
- Linked list: Reverse Linked List, Linked List Cycle
- Trees: Invert Binary Tree, Binary Tree Level Order Traversal, Validate Binary Search Tree, Lowest Common Ancestor of a BST
- Graphs: Number of Islands, Clone Graph, Course Schedule
- Heap and top-k: Top K Frequent Elements, Kth Largest Element in an Array
- Intervals: Merge Intervals, Insert Interval
- Backtracking: Subsets, Combination Sum
- Dynamic programming: House Robber, Coin Change, Unique Paths
Five of those thirty are not in Blind 75 at all: Koko Eating Bananas, Daily Temperatures, Minimum Size Subarray Sum, Subsets and Kth Largest Element in an Array. That is deliberate. Binary search on the answer, monotonic stack, variable-size windows and plain subset generation are four of the most common screen questions, and Blind 75 either skips them or gestures at them once. If your last prep round was Blind 75, do those five first.
What each list under-covers, and what to bolt on
These are family checklists, and every checklist is short somewhere. Knowing where yours is short beats arguing about which list is better.
| List | Shape | What it under-covers | Bolt this on |
|---|---|---|---|
| Blind 75 Grouped by topic heading |
Ten sections, heaviest on trees and dynamic programming, and the section title tells you the technique before you have read the problem | Intervals get five slots, and two of those have historically sat behind LeetCode Premium, so a free account effectively sees three. Tries get three, all filed under "Tree", one of which is a Hard most people skip. There is no monotonic stack, no binary search on the answer, and backtracking shows up twice. | Meeting-room style interval counting, Daily Temperatures, Koko Eating Bananas, Subsets, Permutations. Eight extra problems closes the gap. |
| Grind 75 Ordered by difficulty progression |
Deliberately mixes topics instead of grouping them, and the generator rebuilds the schedule around the weeks and hours you enter | Because it is not grouped, a failure does not tell you which family is weak, so you diagnose that yourself. At short week settings it drops most Hards and thins the graph section to the basics. | Keep a one-line log per problem naming the family you used. That restores the diagnostic the grouping would have handed you. |
| NeetCode 150 Grouped, roughly three problems per pattern |
The Blind 75 families plus tries, backtracking, advanced graphs, greedy, math and bit manipulation as first-class sections | The advanced graph and 2-D dynamic programming sections are heavier than a 45-minute screen justifies. It is also pattern-ordered rather than frequency-ordered, so finishing all 150 still does not guarantee you have met your target company's top twenty. | In the final fortnight, swap the advanced graph section for company-tagged questions and one mock a week. |
Which individual titles are Premium changes over time, so check before you plan a week around a specific problem. The general point survives either way: the shorter list is short somewhere, and interval sweeps plus prefix-tree work are where it is shortest.
Setting the Grind 75 generator honestly
Grind 75 comes with a generator that asks for weeks and hours per week, then rebuilds the schedule to fit. Longer horizons pull in problems beyond the base 75, shorter ones trim towards the highest-value core. Three things go wrong with it.
Enter the hours you will actually do
People with a full-time job type 15 and do 6. The generator then builds a plan they abandon in week 2, which is worse than a smaller plan they finish. Enter last week's real study hours, not the aspirational number.
Set weeks to your interview date minus one
The final week is not for new material. Reserve it for timed 45-minute sessions and the behavioural half, both of which have to be rehearsed out loud rather than read.
Re-run it at the halfway mark
Three problems behind at week 3 of 6 is not a reason to sprint. Regenerate with the weeks you have left and the hours you have proven, and let the tool drop the tail instead of dropping it at random yourself.
When a problem beats the 30-minute cap, a worked explanation beats a bare accepted solution. Code Copilot captures the problem with Ctrl+Alt+S, handles statements spanning several scrolls, and returns the intuition, a dry run and the Big-O, which is the shape of thing you can re-derive a week later. Opening it before the cap is how people finish a list and still fail a screen.
Where none of this helps
A curated list is a coverage instrument. It cannot tell you whether you can talk while you type, and a human is grading precisely that. Two people finish NeetCode 150 and one still fails, because they solved all 150 in silence and the round is a spoken exercise. Book two mocks whichever list you pick, and narrate the whole way through, including the stuck parts.
Lists are also silent on everything that is not algorithms. If your loop includes system design, SQL, a frontend round or a take-home, no amount of NeetCode moves that needle, and the behavioural half often decides between two people who both got the code out. That half can be prepared close to word for word, and the company Q&A generator builds a bank for a named company and role.
One honest note, since a vendor publishes this page. OfferPilot AI does not shorten your list. It is Windows 10 and 11 only, with no macOS, Linux, mobile or browser extension, so on a Mac the answer here is the calendar and the thirty problems and nothing else. Its overlay is excluded from screen capture at the operating-system level, so it does not appear in Zoom, Teams, Meet or a screen recording, and that is a claim about screen sharing alone. It does nothing about a webcam, a lockdown browser, a proctor in the room or a second-device camera, and where an assessment forbids outside help, the honest window for any tool is the weeks before it rather than the hour of it. The free tier is five minutes a day with no card, and the paid tiers sit on the pricing page.
The short version
Under ten days, the thirty. Two to six weeks, Grind 75 with the generator set to hours you will genuinely do. Eight weeks or more, NeetCode 150 for the extra reps per pattern. Whichever row you land on, add the interval sweep and the monotonic stack if you came from Blind 75, and keep the last week clear for talking out loud.