The default recommendation in this category has not shipped a new edition since July 2015. Cracking the Coding Interview is still on its 6th edition — 708 pages, 189 questions, every solution written in Java — and it is still the first title most people name. Eleven years is a long gap in a market where the same loop now adds a system design round for anyone past a couple of years of experience.
We read tables of contents, sample chapters and current print listings for the five books below. We did not measure offer rates, we did not read the C++ or Java editions of Elements of Programming Interviews page by page, and no book here makes anyone pass an interview. The one thing a book does better than a problem list is explain why an approach works. Judge them on that and the shortlist gets short fast.
The expensive mistake is buying two books for the same job
These books cover three jobs that barely overlap: what the process expects of you, drilling problems, and system design. The most common purchase is Cracking the Coding Interview plus Elements of Programming Interviews — two books for job two, nothing for jobs one and three, and about $75 spent on overlapping problem sets.
CtCI's durable value is its non-code half: the walkthroughs of what an interviewer is scoring while you flail, the behavioral prep, the chapter on offers and negotiation. That material aged well. The problems aged less well. Java-only solutions, and a difficulty distribution that sits below what a current senior loop asks. If you interview in Python or Go you translate every solution yourself, which is either useful practice or a tax depending on your patience.
The drill layer: one book is harder than your interview, one is easier
Elements of Programming Interviews in Python carries roughly 250 problems and runs consistently harder than CtCI. The solutions are compact — dense, idiomatic Python with very little hand-holding — which is exactly right if you already solve mediums and punishing if you do not. Its structure is also better for spaced practice than CtCI's: problems are grouped by data structure with explicit variants attached, so you can hammer one shape for an evening instead of bouncing.
Grokking Algorithms, Second Edition (Manning, 2024) is not an interview book at all, and that is the point. Around 350 pages, heavily illustrated, covering binary search, sorting, recursion, hash tables, graphs and Dijkstra, greedy, dynamic programming and k-nearest neighbours, with the second edition adding tree material the first lacked. If a chapter of EPI reads as noise, the failure is upstream of the interview and this is the cheapest fix for it.
Check the language on the cover twice. Elements of Programming Interviews ships as separate Python, Java and C++ books with different ISBNs and near-identical cover art, and the wrong one leaves you with 250 problems in a language you do not interview in. Marketplace listings mix the editions freely. CtCI has no equivalent trap for the opposite reason: there is no Python edition, and the 6th is Java from front to back.
System design, and the one reference that outlives the loop
System Design Interview – An Insider's Guide (Alex Xu, 2020) is 16 chapters of worked designs: rate limiter, consistent hashing, key-value store, unique ID generator, URL shortener, web crawler, notification system, news feed, chat, search autocomplete, YouTube, Google Drive. It is the only book on this page written for the round most candidates lose. Its weakness is structural — it is a set of answers rather than a method, so it is easy to memorise the diagrams and then get taken apart by a single follow-up about write amplification. Read it as worked examples, not as a script.
The Algorithm Design Manual, 3rd edition (Skiena, Springer, 2020) is the outlier: about 800 pages, and Part II is a catalog of roughly 75 classic problems mapped to the approaches that solve them. It is a poor cram book and a good permanent one — the volume you still open two years later when a production problem turns out to be set cover wearing a hat.
Buy in this order, and know when to buy none
Grokking first, but only if you cannot reliably reason through a medium-difficulty problem — every other book on this list assumes a fluency it does not teach. CtCI second, read early enough that its process chapters can still change how you prepare. EPI third as the actual drill. Xu's Volume 1 only if your loop has a design round. Skiena after the offer.
The honest case for buying none: if you already solve mediums in under 30 minutes and can state your complexity without being asked, a book is a worse problem source than a free curated list. Books lose on volume, on feedback, and on recency. They win on explanation. The condition that flips it back toward paper is narrow but real — you keep re-solving the same category of problem and still cannot say why the working approach works.
Keep a review log with three fields per problem: the problem, the wrong first idea you had, and the one sentence that would have saved you. Re-solve on day 3 and day 10. The wrong-first-idea column is the one that pays; it is where the pattern in your own failures becomes visible.
Originally published at pickuma.com. Subscribe to the RSS or follow @pickuma.bsky.social for new reviews.
Top comments (0)