Grind 75 vs Blind 75: reading the successor list against the original

grind75.com doesn’t serve a page of its own any more. I checked it on 20 August 2026 and it answers with a 308 permanent redirect over to techinterviewhandbook.org/grind75, which is where the thing actually lives now. That redirect is a fair summary of the whole Grind 75 vs Blind 75 story. One list got folded into a bigger site, quietly, by the person who wrote both of them.

I read the static pages behind that tool this morning. The about page, the FAQ, the changelog. I wanted to know what the author changed, and more interestingly, what he said his own earlier list got wrong. Everything below is from those pages as they read on 20 August 2026.

Grind 75 vs Blind 75, in one paragraph

Blind 75 is a fixed set of 75 LeetCode problems with no order and no priority attached. Grind 75 is its successor from the same author, Yangshun Tay. It’s a ranked pool of 169 problems that hands you the top 75 by default, then sorts, filters and splits them into weeks based on how many hours you say you can give it.

The site doesn’t hide the lineage. Its title tag ends with “A better Blind 75 you can customize, by the author of Blind 75”, and the meta description on techinterviewhandbook.org/grind75 opens with “Grind 75 is a better version of Blind 75 which goes beyond 75 questions.”

What the author says was wrong with Blind 75

This is the part I found genuinely useful, and it sits on the about page rather than anywhere marketing-shaped. Tay describes himself there as an ex-Meta Staff Engineer and the author of Blind 75. His complaint about his own list is short: “the Blind 75 list is static, non-personalized and questions do not have a priority”. He adds that 75 turned out to be too few for a lot of people, and that a flat list can’t know how much time you’re willing to spend or which topics you want to focus on.

He also says most of the questions are still relevant. Hold on to that. A successor usually shows up carrying a story about how the old thing is obsolete, and this one declines to tell it.

Read the objections again and notice what isn’t in them. He doesn’t say the picks were bad. Everything he lists is about packaging: no priority inside the 75, no way to declare that you’ve got three weeks rather than three months, no way to skip topics your target company doesn’t ask about. Fixing that is worth doing. It’s also a smaller job than the word “successor” makes it sound.

Where the 169 questions came from

The FAQ walks through the construction, and it’s less mysterious than you’d guess. The starting pool came from Blind 75 itself, Sean Prashad’s LeetCode Patterns, questions in Elements of Programming Interviews, and a few other pattern collections. He then sorted LeetCode by popularity, frequency, companies, likes rating and topics covered, and looked at the top 200 in each of those dimensions to spot anything missing. Questions got tagged by pattern, ranked inside their pattern, then flattened into one ordered list and broken into weeks.

Two of the ranking rules stand out. Easy questions go first, because they teach a pattern you’ll reuse later. Hard questions get pushed down, and the stated reason is that they usually run past half an hour while “the most common question difficulty in coding interviews is Medium”. The FAQ is direct about the resulting mix: “You will see that more than half of the questions are of Medium difficulty.”

Premium problems get shoved as far back as possible. The about page lists “No premium questions in the initial list of 75 questions” as a feature, and if you’re not paying LeetCode a monthly fee, that decision will affect your week more than the sort order does.

One credited source worth opening yourself

Sean Prashad’s Leetcode Patterns, which the Grind 75 credits section names, is still up and still free. On 20 August 2026 its header describes it as “A collection of 179 questions grouped by pattern to help you prepare for coding interviews.” and shows 13.6k GitHub stars next to it. Same rough material, sliced by pattern instead of by week. Some people hold that shape in their head better, and it costs nothing to look.

Grind 75 vs Blind 75, side by side

Everything in the Blind 75 column is the author’s own characterisation of his earlier list, since he wrote both. Checked 20 August 2026.

The controls, in the site’s own words

The changelog entries from May 2022 spell out the settings, which helps, because they’re easy to skim past in the interface. Question List is either View all questions or Based on preferences. Order is Difficulty (easy ones first), Topics (important topics earlier) or All Rounded. Grouping is Weeks, Topics or Difficulty. The default Order was switched to Difficulty in May 2022 on the grounds that it seemed more popular, and All Rounded is what you select to get the original ordering back.

Preferences live in the URL, added the same month, so you can bookmark a configured list or send it to a friend. Progress doesn’t. It sits in localStorage. Clear your browser and the ticks are gone, which is a small thing right up until you’re six weeks deep on a work laptop that gets reimaged.

It’s worth knowing why the FAQ pushes you to work the list in order instead of topic by topic. The argument is coverage under a deadline: a real interview can pull from any topic, so a week that touches several patterns beats a week that drills one into the ground. Whether you buy that depends on how far out you are. Three months from a target date, I’d still rather go deep on one weak area at a time and switch to the mixed ordering later.

The old complaint now lands on the successor

Here’s the awkward bit. The most recent entry on the Grind 75 changelog is dated 2022/09/11, and it’s a one-line difficulty tweak to a single problem. So the tool is dynamic in the sense that it re-sorts and re-slices on demand, while the pool it draws from hasn’t visibly moved in a long time. Static was the word used against Blind 75.

I don’t think that’s fatal, and I’d argue the opposite case if pushed. Pattern-based interview questions age slowly, and a pool assembled in 2022 around two pointers, sliding window, graphs, intervals and dynamic programming still maps cleanly onto what a 2026 phone screen throws at you. I could also be wrong about the silence, since a changelog is a promise to log changes rather than proof that nothing happened off the record. But if you’re choosing between the two because one of them sounds newer, the gap is thinner than the framing suggests.

So which list should you actually run?

If there’s a date on your calendar, take Grind 75 and be honest with the hours setting. Overstating your free time is the common failure here, and the weekly split is the main reason to prefer it over a flat list in the first place. Seven hours you’ll really spend beats fifteen you won’t.

If you ground through Blind 75 a couple of years ago, don’t restart it. Push the pool past 75 and work into the remainder of the 169, which is the strongest case the newer list makes for itself.

If you’re early and still shaky on fundamentals, take the Difficulty ordering rather than All Rounded, and accept that you’ll feel slow for about three weeks. Our walkthrough of the Blind 75 list still holds up for that stretch. If you’re weighing a third option, NeetCode 150 against Blind 75 is a genuinely different trade, not a rerun of this one. And for pattern recall rather than raw problem count, the LeetCode patterns cheat sheet is the shorter road.

And if you decide you need neither, that’s a fine answer. These lists earn their keep when your bottleneck is not knowing what to practise next. If your bottleneck is that you can solve things calmly on paper and fall apart against a clock, another sixty problems won’t move it.

Ordering is the problem both lists solve. Retention is the one neither of them touches, and the FAQ says so out loud: “Spaced repetition, also known as distributed practice, is an extremely effective approach for internalizing knowledge into long-term memory.” A list can put questions in a sensible sequence. It can’t drag you back on day nine to redo the one you failed. That part is yours.

Talking the solution out loud is the other gap. You can clear 169 problems in an IDE and still lock up when somebody is watching you type, which is why we build mock interviews at LastRound AI. Two of those in the week before an onsite will tell you more about your real failure mode than the next twenty problems will.

My honest read: the Grind 75 vs Blind 75 decision matters less than the hours you write next to it. Pick whichever ordering you’ll actually follow, then go check whether that changelog has moved past 2022/09/11. If it has, the successor earned its name a second time.