Some interview questions stop being questions and become culture. They show up in books, in tweets, in comedy bits about tech hiring. Engineers remember the first time they heard the manhole-cover question — usually with a kind of secondhand embarrassment, because by the time they got asked, everyone already knew the trick. That recursion is what turns an interview question into a famous one: not whether it is hard, but whether the entire industry has formed an opinion about it.
This hub is a tour of that canon — the brainteasers, coding problems, and behavioral set-pieces that broke out of their original interview rooms and became part of how the industry talks about hiring. Each one has a story: who asked it first, what it was trying to test, why it spread, and what happened when it stopped working. We cover three buckets — brainteasers and logic puzzles, the coding canon, and the behavioral classics — plus the meta-narrative pieces about how the industry got from Microsoft brain-twisters in the 1990s to LeetCode grinding in the 2020s.
Why fame matters in interviews
A famous question is a different beast from a hard question. Hardness is intrinsic to the problem; fame is a property of how the problem moves through the industry. Once a question is famous, two things happen at once. First, candidates over-prepare for it — every prep book, blog post, and YouTube channel covers it, and the answer becomes rehearsed. Second, interviewers stop trusting the signal — a perfect FizzBuzz says nothing about whether someone can write code, only that they read the same book everyone else read. The half-life of a useful question is exactly however long it takes to leak.
That is why the famous canon is mostly a graveyard of questions that used to work. Microsoft’s brainteasers worked in the 1990s when most candidates had never heard of them. Google’s “How many golf balls fit in a school bus” worked until Laszlo Bock looked at the data and decided it was a worse predictor than just looking at the resume. Whiteboard coding worked until Max Howell’s tweet exposed how arbitrary it could be. The questions did not get easier — the industry got better at preparing for them, and that was the same thing as them getting useless.
1. Brainteasers, logic puzzles, and Fermi estimation
This is the oldest tradition in tech interviewing, and the most thoroughly retired. The Microsoft era of the 1990s — captured in the book How Would You Move Mount Fuji? — leaned hard into puzzles like the manhole cover, the river crossings, the burning ropes, and the shrunk-to-the-size-of-a-pencil scenario. The premise was that they tested raw problem-solving under pressure. The reality, decades later, is that they mostly tested whether candidates had encountered the puzzle before.
Wall Street took the same tradition and never stopped. Trading firms, hedge funds, and quant prop shops still ask probability and game-theory puzzles every day — the ant on a clockface, the pirates dividing gold, the two-children paradox, fair-coin-from-biased, the 100 prisoners and the light bulb. The difference is that on Wall Street, the puzzles double as a test of probability fluency, which is a job-relevant skill for a quant trader in a way that “How would you move Mount Fuji” never was for a Microsoft PM.
Famous brainteasers covered in this section: the manhole cover, Mount Fuji, the blender / shrunk-to-a-pencil, golf balls in a school bus, piano tuners in Chicago, weighing a 747, the ant on a clockface (Goldman Sachs), pirates dividing gold, the two-children paradox, fair-coin-from-biased, coin flip streaks (HT vs TT), the 100 prisoners and the light bulb, the secretary problem (1/e rule), and the two-envelopes paradox.
2. The coding canon
Coding interviews have their own famous list, and it splits cleanly into the cultural and the canonical. The cultural ones are famous because of what happened around them — FizzBuzz became a meme because Imran Ghory’s 2007 blog post claimed most CS graduates could not solve it, and the internet has been arguing about that claim ever since. Invert a binary tree became famous because Max Howell, who wrote Homebrew, tweeted that Google rejected him for not being able to do it on a whiteboard, and that single tweet did more to shift industry opinion against whiteboard hiring than any academic paper ever could.
The canonical ones are famous because they are the prototypes of entire problem patterns. Trapping rain water is the optimization-journey poster child — the brute force is O(n²), the optimal is O(n) with two pointers, and walking from one to the other is exactly the kind of thinking the interview is trying to elicit. Median of two sorted arrays is the nastiest binary search you will ever do, and once you can solve it cleanly, every other binary search on answer-space feels easy. Largest rectangle in histogram is where you learn what a monotonic stack is for. These problems are famous not because of a tweet but because they teach a generalizable technique, and once you know them, you see them everywhere.
The system-design canon is shorter and younger. Design an LRU cache is the gateway drug — small enough to code in 20 minutes, but it forces you to think about data structures and access patterns together. Design Twitter, design a URL shortener, and design a rate limiter are the architecture-thinking poster children, the questions that signal whether a candidate can scope an ambiguous prompt and reason about trade-offs at scale.
3. The behavioral canon
The behavioral canon is the most resilient of the three because the questions are job-relevant in a way the brainteasers never were. “Tell me about yourself” is the universal opener — it is asked literally everywhere, it tests a real skill (can this person frame their own narrative), and even though every candidate prepares for it, the bad answers are still very bad. “Tell me about a time you failed” demands the most from STAR — the candidate has to tell a story that is genuinely a failure (not a humblebrag), explain what they learned, and convince the interviewer they would handle it differently now.
Then there are the company-specific frameworks. Amazon’s 16 Leadership Principles are the most-searched behavioral-prep keyword in tech because Amazon hires aggressively, asks the same set of questions across all its loops, and publishes the LPs openly. Google’s “Googleyness” round, after the post-2018 culture shift, became its own thing. The “pen-selling” question from Wolf of Wall Street still gets asked in sales-adjacent loops — and what it actually tests is not very different from what a behavioral conflict-resolution question tests.
4. Meta-narrative pieces
Some pieces in this section are not about a single question but about the arc of how the industry got to where it is. The Max Howell effect — one tweet, one shifted Overton window. Google’s 2013 brainteaser mea culpa, when Laszlo Bock told the New York Times that “brainteasers are a complete waste of time” because the data showed they predicted nothing. The Cracking the Coding Interview effect — how Gayle Laakmann McDowell’s book standardized the entire industry’s expectation of what a coding interview looks like. The 30-year arc from brainteasers to LeetCode to take-homes, and what each shift was reacting against.
The Lyft frontend debate from 2024 — when a senior frontend engineer publicly criticized Lyft for asking him a hard graph algorithm in a UI role — is the most recent flashpoint. It reopened the same argument the Max Howell tweet started: what does the question actually predict, and how much of the answer is just rehearsal?
What is still asked in 2026
The honest answer is: most of the canon is dead in tech, alive in finance, and remembered everywhere. Big tech (FAANG, AI labs, fintech tier 1) has mostly settled on LeetCode-style mediums and hards plus system design and behavioral. Pure brainteasers are gone. Estimation questions still appear in product manager interviews and in some staff-engineer rounds. Wall Street still asks probability puzzles every day, and the canon there is alive and well — Jane Street, Citadel, HRT, Two Sigma, and the others all draw from the same pool of probability classics.
What is universal: every candidate still gets “tell me about yourself”, every senior candidate gets “tell me about a time you failed”, and every tech candidate over staff level gets at least one architectural-thinking system design question. Those three never went out of fashion because they actually predict something.
How to use this section
For each famous question, we cover four things: the cultural story (who asked it first and why it spread), the technical answer or framework (what a good response actually looks like), what the question is really probing (often different from what the question literally asks), and whether it is still in the live rotation in 2026. The goal is not to give you a script — the questions that have a script are the ones that no longer work. The goal is to give you the historical context, so when one of these questions comes up, you understand what kind of signal the interviewer is trying to extract, and you can answer the underlying question rather than the surface one.
Start with FizzBuzz and the Max Howell binary-tree story if you want the modern coding canon, the manhole cover and Mount Fuji if you want the dead Microsoft tradition, the ant-on-a-clockface and the two-children paradox if you want Wall Street, and the STAR method and Amazon LPs if you want behavioral. The pieces cross-link to the technical solutions where they exist on this site, so you can flip from cultural context into the specific algorithm or framework as needed.
Frequently Asked Questions
Are brainteasers still asked at FAANG in 2026?
Almost never. Google publicly retired them in 2013 after Laszlo Bock said the data did not support them. Microsoft phased them out across the 2000s and 2010s. Meta, Amazon, and Apple were never as into them as Microsoft was. The exception is product-manager loops, where Fermi-style estimation questions still appear, and some staff-engineer rounds where ambiguous problem-framing questions resemble brainteasers.
What about Wall Street and quant firms?
Probability brainteasers are core to quant interviews — Jane Street, Citadel, HRT, Two Sigma, DRW, Optiver, SIG, Akuna, IMC, Jump, and Tower all ask them. The questions are job-relevant there in a way they never were in tech: a quant trader actually does need to compute expected values quickly under pressure.
What is the most famous interview question of all time?
Probably the manhole cover for the Microsoft era, FizzBuzz for the modern coding era, and the ant on a clockface for Wall Street. The manhole cover is the most-cited brainteaser in industry retrospectives. FizzBuzz is the only interview question that became a meme outside of tech.
Is whiteboard coding dead?
Mostly. Most onsites now use a collaborative editor (CoderPad, CodeSignal, Google Docs, or just a shared screen). Pure whiteboard coding survives in some teams at Microsoft, Apple, and a few legacy tech firms, but it is no longer the default. The Max Howell tweet accelerated this shift, but it was already happening.
How do I prepare for famous questions specifically?
Read the cultural context, then practice the underlying skill. The famous questions are rehearsed enough that a memorized answer reads as memorized — what differentiates is whether you understand the technique well enough to handle a variation. For FizzBuzz, the variation is “do it without modulo” or “make it scale to a billion numbers”. For invert a binary tree, the variation is “do it iteratively” or “what if the tree is enormous and does not fit in memory”. Knowing the answer is table stakes; being able to discuss the trade-offs is the actual signal.