On June 19, 2013, Adam Bryant of the New York Times published an interview with Laszlo Bock, then Google‘s Senior Vice President of People Operations. In the middle of a conversation about how Google was using internal data to overhaul its hiring, Bock dropped a sentence that became the most-cited mea culpa in tech-hiring history:
“On the hiring side, we found that brainteasers are a complete waste of time. How many golf balls can you fit into an airplane? How many gas stations in Manhattan? A complete waste of time. They don’t predict anything. They serve primarily to make the interviewer feel smart.”
The line spread through tech-industry channels within hours. Within weeks it had been republished, summarized, and quoted in every “how to interview” article on the internet. Within months, brainteasers as an interview format were on a clear downward slope across the industry. Bock’s statement did not single-handedly kill the genre — Microsoft and others had been quietly de-emphasizing brainteasers for years — but it provided the public, named, data-backed retreat that made the change visible and accelerated it.
Why Google specifically
Google’s adoption of brainteasers in the 2000s had been so prominent that the company was synonymous with the format. The “How many piano tuners are there in Chicago?” question, the “How many golf balls fit in a school bus?” question, and various Fermi estimation problems were core to Google’s interview reputation. Books like William Poundstone’s Are You Smart Enough to Work at Google? (2012) had cemented the association in the public mind.
So when Google’s most senior people-operations executive said publicly that the questions did not predict job performance, the statement carried unusual weight. It was not a critic dismissing a practice from outside; it was the institution’s own SVP saying the institution had been doing it wrong. That self-disclosure structure is what made the 2013 statement land harder than any external critique could have.
The data Google looked at
Bock’s claim was based on internal Google research that the company had been doing on its own hiring outcomes. Google’s People Analytics team, led by Prasad Setty during this period, had been correlating interview performance with on-the-job performance for years. The headline findings, as Bock and others described them across multiple interviews and Bock’s later book Work Rules! (2015):
- Brainteaser performance had near-zero correlation with on-the-job performance. Candidates who did well on the puzzles did not perform measurably better than candidates who did poorly.
- Structured behavioral interviews predicted performance much better. Interviewers who used standardized rubrics and asked the same questions across candidates produced more reliable signal.
- Work-sample tests (where candidates do the actual work in front of an interviewer) were the single best predictor. Coding interviews fall in this bucket and survived the analysis.
- The number of interviewers required to make a hire/no-hire decision was much lower than Google had been using. Going from a 12-interviewer loop to a 4-interviewer loop captured most of the predictive value with much less time investment.
The brainteaser conclusion was the most quotable, but the broader reform Bock was pushing was about structured interviewing in general — same questions, same rubric, same scoring scale. The brainteasers were the most visible casualty of that reform, but they were not the whole reform.
What replaced the brainteasers
By 2013, Google’s interview structure had already started shifting away from brainteasers and toward four core question types: cognitive ability tests (in the form of coding problems and structured analysis), conscientiousness tests (in the form of behavioral interviews about past work), leadership tests, and “Googleyness” (a culture-fit dimension). The brainteasers had occupied the cognitive-ability slot, and they were replaced by coding problems and structured analytical reasoning, which the data showed predicted better.
The interview format that emerged at Google in the years after 2013 looked roughly like:
- 2 coding rounds (medium and hard LeetCode-tier problems)
- 1 system design round (for senior+ roles)
- 1 “general cognitive ability” round, replaced over time with more coding
- 1 leadership/Googleyness round (structured behavioral)
This is roughly the shape that has stabilized at most FAANG companies today. The brainteaser slot was definitively eliminated, and the time was reallocated to coding and behavioral.
The industry effect
The 2013 statement traveled fast because it was specific, public, and authoritative. Within a year, other large tech companies were quietly reviewing their brainteaser usage. Within three years, brainteasers had been substantially phased out across FAANG, AI labs, and tier-1 startups. By 2020, asking a brainteaser earnestly in a tech interview was considered a sign of an outdated process.
The effect was not uniform. Microsoft, where brainteasers had originated, retreated more slowly and more quietly — there was no parallel “Microsoft mea culpa” public statement, but the practice was gone from most Microsoft loops by the late 2010s. Some smaller tech companies kept the format longer because they were not measuring outcomes at the scale Google was. And critically, Wall Street and the quant industry never adopted Bock’s reasoning — their probability puzzles continued to be asked because the underlying skill was actually job-relevant for traders, in a way that “how many golf balls fit in a school bus” never was for software engineers.
The deeper argument
Bock’s statement is sometimes summarized as “brainteasers are bad”. The actual claim was more specific and more interesting: brainteasers do not predict on-the-job performance, but only because the underlying skill (loose problem-solving on novel domains) is mismatched with the actual job (writing software in a structured codebase). The same Google research showed that structured analytical reasoning, applied to job-relevant problems, did predict performance.
The right reading of the 2013 mea culpa is not “puzzles are bad in interviews” but “use puzzles whose underlying skill matches the job”. Coding interviews are puzzles. System design interviews are puzzles. The Wall Street probability questions are puzzles. The difference is that those puzzles draw from the same skill pool the job draws from. Brainteasers about manhole covers and golf balls do not, and that is the part that did not predict.
What this means for candidates today
Three implications for anyone preparing for tech interviews in 2026:
- You can ignore the brainteaser tradition entirely. The questions are essentially extinct in tech. Books from the 2000s that emphasize them are out of date.
- Structured behavioral interviews are real. The questions in those rounds are deliberately scripted, scored against rubrics, and normalized across candidates. Treating them as throwaway is a serious error.
- The work-sample principle generalizes. If an interview format is asking you to do something that resembles the actual job, take it seriously — coding, system design, take-home assignments, live debugging. If a format is asking you to do something that does not resemble the job, the company is probably not measuring outcomes carefully, and the signal is correspondingly noisy.
The book
Bock expanded the 2013 statements into a full book in 2015: Work Rules! Insights from Inside Google That Will Transform How You Live and Lead. The book is the canonical source for how Google’s people-analytics-driven approach to hiring developed, and it is widely read in HR and engineering management circles. The brainteaser mea culpa is a small part of the book; most of it is about structured interviewing, performance management, and compensation. But the brainteaser line is what travels, because it has the satisfying shape of a public retreat from a famous practice.
Frequently Asked Questions
What did Laszlo Bock actually say?
“Brainteasers are a complete waste of time. They don’t predict anything. They serve primarily to make the interviewer feel smart.” From the 2013 New York Times interview with Adam Bryant.
Did Google immediately stop asking brainteasers?
The transition was already underway by 2013 and continued for several more years. Bock’s statement was both a public acknowledgment of the change and an accelerant. By 2016 or so, brainteasers were essentially extinct at Google.
What is the data behind the claim?
Google’s internal People Analytics team correlated interview-round scores with on-the-job performance reviews and promotion outcomes for years. The brainteaser rounds showed near-zero correlation. Structured behavioral interviews and work-sample tests (coding) showed strong correlation.
Are brainteasers gone everywhere?
Mostly gone in tech. Still alive and well in finance and quant interviewing, where probability puzzles correlate with the actual skills traders need. Still occasionally seen in PM interviews as Fermi estimation, where market sizing is a job-relevant skill.
What replaced brainteasers at Google?
Coding interviews (LeetCode-style mediums and hards), system design interviews, and structured behavioral rounds with explicit rubrics. The total number of interview rounds also went down — Google had been using up to 12 rounds for some candidates, and reduced this to roughly 4 after the analytics work showed diminishing returns.