Sep. 28–Oct. 5 interview digest: accused of cheating for solving too fast, and Meta's median jumps $27K
This week's reports turn a cheating accusation, an Apple loop that went silent, a MongoDB role frozen overnight, and a $27K jump in Meta's median into drills and negotiation references.
The seven days ending Oct. 5, 2026 produced an accusation where the week before had produced a question. A candidate solved the coding problem in an Amazon round, began explaining his approach, and heard the interviewer ask whether he had used outside tools or looked it up beforehand. He said he had not. The interviewer pressed: "it just came together very quickly, I want to make sure it's your own work." The candidate wrote that he had seen the same pattern before, that preparation was exactly what had made him fast, and that being prepared had started to read as suspicious. He spent the rest of the round over-explaining every line to prove the work was his, which he said made him look more nervous. "That was how I blew this interview." 1
The post drew 23 likes and 23 comments in its first two days. The replies split the way this topic now splits. One commenter wrote that other candidates cheating with an AI tool had ruined it for everyone. An Amazon employee answered that a candidate who pulls a finished solution out without showing the problem-solving behind it will fail the round whoever wrote the code, and advised walking the interviewer through the approach before typing. 1
The same week, the money moved. Levels.fyi refreshed its four software-engineer pages on Oct. 5, and Meta's median total compensation rose to $482K from $455K a week earlier, while Google's fell to $285K from $301K. 23
Glassdoor's report pages were readable this window for Google, Microsoft and Apple; Meta and Amazon returned the Cloudflare security page, so this issue has no dated Meta or Amazon report in it. Blind supplied first-person accounts with relative age markers (
2d, 4d, 5d) and a like, comment and view counter under each post, read off page screenshots. Reddit's interview-discussion thread and its offer threads returned verified scores but no current-window interview report. Four limits hold across the issue: no source publishes a comparable vote total across platforms, so the two report tables below are ordered inside their own platform and their counters never combine; Glassdoor prints no engagement figure on an individual report, so its rows carry a date and an outcome instead; Glassdoor's rows record the posting date while the interview itself may have run months earlier, and the interview month is printed beside the date where the report gives one; and nobody publishes the applicant volumes and completed-interview counts a company-wide pass rate would need, so this issue reports individual outcomes together with the conditions that produced them.The week in one read
- An Amazon candidate was accused of cheating for solving too fast. He solved the problem and was asked whether he had looked it up beforehand. He wrote that the accusation, and the over-explaining it triggered, is what ended the round. 1
- An Apple loop finished and then went quiet. The poster completed a four-round Apple Austin loop on Sep. 18, felt the coding rounds had gone well, followed up with the recruiter on Sep. 22, and got no reply. The post reached 6,587 views, the largest audience of any interview report this window, and its 42 comments describe waits of six weeks and being kept as a backup. 4
- An Amazon L6 screen scheduled for an hour ended after about 25 minutes. The poster asked whether anyone had reached a loop after a hiring-manager conversation that short. The 31 comments split between reading it as a rejection and reading it as a manager who had already decided. 5
- A MongoDB loop went well and the role was frozen the next day. The recruiter said the position had been frozen and the posting came down; the poster asked whether rounds already cleared carry to another team or have to be repeated. 6
- Two OpenAI questions got names. A poster found a coding question titled "Alert Routing with Nested Routes" and a system-design question titled "Online competitive typing speed game" newly added to a paid question bank, and asked who had been asked them. 7
- A LinkedIn staff candidate was judged "high senior." The AI-coding round ran regular coding with several subproblems; he finished three, and the interviewer wanted a math problem instead of the fourth. His read is a down-level. He wrote that his total compensation is around $600K now and drops toward $400K in February when his vesting cliff hits. 8
Evidence at a glance
Loading stats card…
The nine Glassdoor reports are three each at Google, Microsoft and Apple, all posted between Sep. 29 and Oct. 5. Company aggregates on the same pages: Google lists 26,172 interviews, 63.7% positive experience and an average 38 days to hire 9; Microsoft's software-engineer page lists 1,722 interviews, 64% positive and 29 days 10; Apple lists 15,160 interviews, 64.4% positive and 31 days 11.
Report cards
On Blind, ordered by the view counter
Blind shows a relative age rather than an absolute date, and that age is reproduced here exactly as displayed. The counters arrived on different days, so a
5d post has had longer to accumulate views than a 2d one.| Company or scope | Freshness shown | Round, task, or process | Reported outcome | Engagement and evidence boundary |
|---|---|---|---|---|
| Apple | 5d | Austin IS&T software engineer (ICT3). A four-round loop on Sep. 14, 15, 17 and 18: coding, technical, behavioral/manager and system design. The poster solved the coding questions and felt the interviews went well, then followed up with the recruiter on Sep. 22. | None reported; awaiting a decision. | 4 likes, 42 comments, 6,587 views. Blind's generated summary of the 42 comments, dated Oct. 5, records commenters waiting six weeks and others saying Apple keeps several candidates in play until it decides. The poster is tagged Morgan Stanley. 4 |
| Amazon | 4d | An L6 hiring-manager screen booked for an hour ran about 25 minutes. | None reported; the poster asked whether anyone had reached a loop after a manager conversation that short. | 6 likes, 31 comments, 3,065 views. Blind's summary of 10 comments, dated Oct. 5, splits between "it's a reject" and commenters saying a manager who ends early has usually decided. A Google employee notes Google hiring-manager rounds run about 30 minutes by design. 5 |
| OpenAI | 2d | Two questions the poster found newly added to a paid question bank: a coding question named "Alert Routing with Nested Routes" and a system-design question named "Online competitive typing speed game". | None reported; the poster is asking who has been asked them and what they cover. | 5 likes, 17 comments, 2,553 views. Commenters call the question bank a scam that sells outdated questions; one says an unrelated online-chess design question shares the typing-game's shape. The questions are reported secondhand from a paid bank, not from a candidate who was asked them. 7 |
| Amazon | 2d | One coding round: the candidate solved the problem, started explaining, and was asked whether he had used outside tools or looked it up beforehand. | None reported; the poster writes that the round is how he blew the interview. | 23 likes, 23 comments, 1,493 views. The poster is tagged New; the post sits in Blind's Tech Industry channel and names no level. Replies split between "other cheaters ruined it for you" and advice to narrate the solution before typing. 1 |
2d | Staff AI-coding round: regular coding with multiple subproblems. The poster finished three of four and was blocked from the fourth by a math problem. One round ran 15 minutes with no follow-ups; the craftsman round kept repeating "craft" without defining it; coding went well. | Verdict: "high senior". The poster reads it as a down-level and is weighing the jump. | 1 like, 20 comments, 905 views. The poster's own total compensation is given as E5 "somewhere in 600K", dropping to about 400K after February when the vesting cliff hits. Commenters warn a senior offer would pay staff-level work at roughly 60% of the compensation. 8 | |
| Apple | 4d | An upcoming Apple coding interview reported as focused on Go, especially concurrency. | None reported; the poster is asking which problem patterns to expect. | 0 likes, 7 comments, 204 views. One commenter says they had the same round; the poster replies that they messaged that commenter directly, so the round's content stayed out of the thread. 12 |
| MongoDB | 2d | A technical loop for a senior software engineer role. The design round switched interviewer at the last minute and turned conversational; the other rounds went well. The recruiter said the next day that the position had been frozen and would share the profile with other teams. | Loop completed, no debrief; the poster asks whether cleared rounds carry over. | 0 likes, 1 comment, 129 views. The poster checked the job posting and found it taken down; a six-participant poll under the post asks whether cleared rounds carry over. 6 |
| Capital One | 3d | Senior lead software engineer "power day", four rounds. A case on an AI feature with a code-review dry run and a live modification; system design for a banking application with deposits, withdrawals and transfers; behavioral questions, three on projects and cross-team work and the fourth on AI use; and an object-oriented Trie problem with three levels of functionality added in sequence. | Reportedly strong through three rounds, ran out of time on the coding round and expects a rejection. | 1 like, 7 comments, 120 views. The poster reports 10 years of experience and no other loop in flight. 13 |
4h | An L6 phone screen: coding and a Googleyness round completed, more than a week without an answer. | None reported; the poster is asking how long Google takes. | 0 likes, 0 comments, 67 views, four hours after posting. The post is a single question with no replies yet, so it supports no judgment about the process. 14 | |
| Anthropic vs. Google | 3d | An offer comparison rather than a loop: Google L5 at $520K total compensation (base $245K, 20% bonus) against an Anthropic L5 at $575K (base $315K). | Both offers in hand; the poster is asking whether the Anthropic offer can be used as leverage. | 28 likes, 67 comments, 6,762 views. Commenters suggest the Anthropic offer may itself be a down-level, argue its compensation is fixed, and split on whether the lab's equity is worth the risk. The poster reports 7 years of experience. 15 |
| Apple | 4d | An offer evaluation: a hardware role, an ICT5 offer at a $210K base, after a full loop. The recruiter said the compensation team would not consider the poster's roughly $800K of unvested paper equity, and suggested staying at the current employer. | Offer received; the poster is deciding whether to ask for a best-and-final. | 11 likes, 64 comments, about 11,000 views. The poster reports a PhD and 12 years of experience, and that the hiring manager remained interested after the recruiter's reply. 16 |
On Glassdoor, ordered by report date
Glassdoor publishes the date the report was posted, the candidate's outcome, and the candidate's own ratings for difficulty and experience. It publishes no per-report engagement figure. Where a report gives the month it was interviewed, that month is printed beside the posting date.
| Company | Report date | Round, task, or process | Reported outcome | Ratings attached |
|---|---|---|---|---|
| Apple | Oct 5, 2026 · interviewed Aug 2026 | Cupertino, CA. A recruiter screen that asked a few JavaScript concepts and how the candidate uses AI for productivity; a technical screen; then a five-stage virtual loop — hiring manager on full-stack concepts and JavaScript coding, three LeetCode problems (one easy, two medium), two more LeetCode problems, a system-design round that opened by clarifying requirements and constraints, and a behavioral round the recruiter had flagged as communication-heavy. | No offer | Average difficulty; positive experience 17 |
| Oct 4, 2026 | Pittsburgh, PA. One coding round asking for a hard variant of tic-tac-toe, explained fully and coded about 75% before time ran out; and one behavioral round, including a question about a teammate who takes credit for shared work. | No offer | Difficult; negative experience 18 | |
| Oct 4, 2026 | Toronto, ON. The questions were described as general software problem solving; the system-design round was described as standard but varying by team. | Declined offer | Average difficulty; positive experience 19 | |
| Oct 4, 2026 | London, England. The candidate described the process shape: a recruiter screen, one or more technical phone or video rounds on coding and role knowledge, then a virtual loop of coding, system design and Googleyness before a hiring committee. | Declined offer | Difficult; positive experience 20 | |
| Apple | Oct 3, 2026 · interviewed Feb 2026 | An SDE intern loop, run team by team: one LeetCode round with two medium questions, one backend round on API and backend design, and a manager round of behavioral and case-based questions. The report notes a strong preference for a strong résumé and college background. | No offer | Average difficulty; positive experience 21 |
| Apple | Oct 3, 2026 | San Francisco, CA. A hardware role: four rounds with a hiring manager, a recruiter and a team lead, then a day of seven interviews across team members, covering C++, Python, sensor design and signal processing. | No offer | Difficult; neutral experience 22 |
| Microsoft | Oct 1, 2026 · interviewed Sep 2026 | Redmond, WA. A recruiter call and résumé walkthrough; a HackerRank assessment of three to four questions in 90 minutes; a 45-minute phone screen that opened on one medium arrays question and moved to object-oriented and database questions; then a four-round onsite in a single day — two coding rounds on trees and strings, a system-design round on a distributed message queue, and a behavioral round with a hiring manager. The report notes the code has to compile and handle edge cases, and that one interviewer spent 15 to 20 minutes on résumé projects before giving the question. | Accepted offer | Average difficulty; positive experience 23 |
| Microsoft | Sep 30, 2026 · interviewed Feb 2025 | Bucharest, Romania. Three technical interviews on algorithmic problem solving, easy to medium, plus behavioral questions on team fit and past roles. | Accepted offer | Average difficulty; neutral experience 24 |
| Microsoft | Sep 29, 2026 | Prague. An employee referral led to a behavioral screen with a principal manager, then a loop of system design, an "AA" (as-appropriate) round and coding, each about 45 minutes. | No offer | Average difficulty; neutral experience 25 |
The question bank you can practice now
1. Code that has to run, in a four-round day
Microsoft's Redmond report lays out a full onsite compressed into one day: two coding rounds on trees and strings, a system-design round on a distributed message queue, and a behavioral round with a hiring manager. The report adds a detail that changes how to practice: the code has to compile and clear edge cases, so a solution that only traces correctly on paper does not count. The assessment before it held three to four questions in 90 minutes. 23
A drill that matches the two clocks:
- Set a 30-minute timer, write one tree problem and one string problem end to end, and run the code before the timer ends. Repeat until edge cases pass on the first run, not the third.
- In 40 minutes, sketch one message-queue design: producers, partitions, consumer groups, delivery guarantees, and what happens when a consumer falls behind. Say out loud which guarantee you chose and what it costs.
2. Apple's five-stage loop and the communication bar
Apple's Cupertino report sets out five stages after the recruiter screen: a hiring-manager round on full-stack concepts, two LeetCode-heavy technical rounds, a system-design round, and a behavioral round. Two details are worth copying into prep. The first is the recruiter screen itself, which asked a few JavaScript concepts and how the candidate uses AI for productivity. The second is the reason the loop exists as five rounds: the recruiter told the candidate the team had high expectations on communication, and the behavioral round tested exactly that. 17
The Apple intern report from two days earlier shows the same three-part shape at a smaller scale, one LeetCode round, one backend round on API design, and a manager round of behavioral and case questions, and notes that the résumé and school carry weight before the loop even starts. 21
A drill for the communication bar:
- Take one finished project and deliver a 90-second version, then a 30-second version, then answer "what would you change now?" in two sentences. The Apple report's behavioral round rewards that compression more than a long story.
3. The two OpenAI questions now circulating
The names to watch are "Alert Routing with Nested Routes" for coding and "Online competitive typing speed game" for system design. 7 The names are worth having. The poster found them in a paid question bank, and commenters under the post call that bank a scam that resells outdated questions; the thread contains no candidate who was asked either question in an OpenAI loop. What the names still give you is the shape. Routing alerts through nested route hierarchies is a tree-plus-matching problem, and a competitive typing game is a real-time system problem: rooms, matchmaking, per-keystroke progress, and a leaderboard that stays consistent under load. Both are worth a whiteboard hour even if the leak is worthless.
4. Apple's Go round
An Apple candidate heading into a coding interview was told it would focus on Go, especially concurrency, and asked what to expect. 12 The thread produced no question list, so the useful work happens before the round: be able to write a worker pool with a bounded channel, cancel a context mid-flight, protect shared state with a mutex against a channel, and explain when you would choose each. Those four moves cover most Go concurrency prompts, and this report confirms Go is a live target at Apple at least once this window.
5. Don't let speed read as cheating
The Amazon candidate gave a correct answer fast enough to look copied, in a season when interviewers are watching for exactly that. 1 The countermeasure comes from the replies. The Amazon employee's advice is to narrate before typing: state the brute-force approach, its complexity, the optimization you rejected, and only then write the optimal solution. An interviewer who has heard the reasoning has no reason to ask whether the code is yours.
A drill:
- Solve one problem you have already seen, out loud, to a recording. Counting your own narration-to-typing ratio tells you whether a fast solve reads as prepared or as suspicious. If you type first and explain after, change the order.
The AI question moved from the interviewer's side to the candidate's
Last week's reports described suspicion running outward: an interviewer explaining how he catches candidates using AI mid-round. This week it ran inward. The Amazon candidate was judged, in effect, for looking like a tool user. 1
The same question is now standard furniture in loops across several companies. Capital One's behavioral round asked four questions, three about projects and cross-team work and the fourth about the candidate's use of AI. 13 Apple's recruiter screen asked how the candidate uses AI for productivity. 17 And the questions themselves are leaking into paid banks fast enough that a named OpenAI coding and system-design question can appear on Blind within days, with the bank's credibility already in dispute. 7
The practical answer to all of it is the same one the Amazon thread arrived at. The loop is testing whether you can stand behind each line, so practice saying why a line exists, what you tried first, and what you would change under a new constraint. That skill survives a cheat-suspicion round, a "how do you use AI" question, and an AI-assisted round alike, and it is the part of the loop you can rehearse alone this week.
After the loop: the freeze, the ghost, and the down-level
Four reports this window describe what happens after the interviews are over, and none of it is in the candidate's hands.
MongoDB's is the fastest reversal. A senior software engineer candidate went through the technical loop, felt it had gone well, and heard from the recruiter the next day that the position was frozen; the posting came down. The candidate's question, whether rounds already cleared carry over to another team or have to be repeated, is the one to ask in writing the moment a freeze is mentioned. 6
Apple's is the slowest. The Austin ICT3 candidate finished a four-round loop on Sep. 18, followed up on Sep. 22, and heard nothing. Under the post, commenters describe waiting six weeks and conclude that Apple keeps several candidates warm until it picks one. The candidate's own question — whether the posting going inactive means anything — drew no firm answer, which is its own signal about how much a candidate can read into silence. 4
The Amazon L6 screen is the shortest. An hour was booked; the conversation ran about 25 minutes. The comments split between rejection and a manager who had already decided, and one Google employee noted that Google's own hiring-manager rounds are scheduled for roughly 30 minutes, which makes an early end at Amazon less legible than it looks. 5
LinkedIn's is the one that turned into a verdict. The staff candidate was rated "high senior" after a loop where the AI-coding round moved the goalposts to a math problem and the craftsman round never defined what it was assessing. With his cliff approaching in February, the decision is whether a title without the compensation is worth the move. The replies lean toward treating the verdict as a down-level: one LinkedIn employee wrote not to take a senior offer, arguing it means working at staff level for far less. 8
The pattern across all four is the same. Each of these four outcomes followed from a hiring decision made elsewhere, once the candidate had finished interviewing. The candidate's leverage is limited to asking early and in writing — whether a freeze voids cleared rounds, what a "high senior" verdict means for level and compensation, and when a decision is expected — so that a silence or a freeze arrives with terms already on the table.
The money this week
Levels.fyi refreshed all four software-engineer pages on Oct. 5. Meta and Amazon moved up week over week; Google and Stripe moved down.
Loading chart…
These are platform medians from self-reported, verified submissions. They describe the market; a single offer is built from its own base, bonus and equity. The level detail moved with them. Google's L5 median sits at $451K and its L6 at $756K 3; Meta's E5 is $454K and its E6 $747K 2; Amazon's L6 SDE III is $384K and its L7 $642K 26; Stripe's L3 is $446K and its L4 $663K 27.
Two first-person offer threads this window show how far a single number can be from the median.
The Anthropic-vs-Google poster expected a lab premium and got a $55K gap. His Google L5 offer stands at $520K total compensation with a $245K base and a 20% bonus target; his Anthropic L5 offer is $575K with a $315K base. Commenters pushed back on the premise, suggesting the Anthropic level may correspond to an L4 elsewhere and that Anthropic's compensation is largely fixed. Others argued the equity is the real variable. 15
The Apple hardware poster is the harder case to price. He holds an ICT5 offer with a $210K base, and about $800K of unvested equity at a private company that he expects to go public next year. The recruiter told him the compensation team would not consider the paper equity and suggested he stay where he is. That reply is the negotiation content: a private company's paper equity is worth nothing to a recruiter until it is liquid, and a candidate holding only paper has little to trade. The counter-move the thread converges on is to get a competing, cash-backed offer before asking Apple for a best-and-final. 16
Reddit supplied no interview report this window, so it appears here as background only. The subreddit's highest-scoring threads were job-market and AI-workplace discussions, none tied to a specific loop: "This comment summarizes the current state of SWE the most accurately" at 535 points and 265 comments, and "Is anyone else's company having second thoughts about AI?" at 168 points and 89 comments. 2829 The one offer thread with numbers was a new graduate in Toronto comparing a current role at CAD $140K plus stock against a Stripe offer at CAD $123K base, $48K stock a year, a $25K signing bonus and a $13K relocation stipend. 30
Before the next loop
Three things this week's reports make worth doing before the next interview.
Rehearse the narration first. Speed now invites a cheating question, and the fix costs nothing but a change of order: say the approach and the rejected alternative before you type, so the interviewer has heard the reasoning before seeing the code. 1
Ask for the terms of a silence, in writing. Whether frozen roles void cleared rounds, what a verdict like "high senior" means for level and compensation, and when a decision is expected are all questions that cost nothing to ask and that a MongoDB-style freeze or an Apple-style ghost makes urgent after the fact. 68
Separate the platform median from your offer. Levels.fyi's medians moved $27K up at Meta and $16K down at Google in one week, while two candidates this window were negotiating gaps of $55K and a refused $800K of paper. Check the median for context and then price your own offer from its own base, bonus and equity, with the vesting schedule in front of you. 23
References
- 1
- 2Meta Software Engineer Salaries
levels.fyi
- 3Google Software Engineer Salaries
levels.fyi
- 4Apple Interview
teamblind.com
- 525 min Interview at Amazon L6 Role
teamblind.com
- 6MongoDB interview loop went well but hiring freeze
teamblind.com
- 7OpenAI added two new interview questions
teamblind.com
- 8LinkedIn staff down leveled
teamblind.com
- 9Google interview questions
glassdoor.com
- 10Microsoft Software Engineer interview questions
glassdoor.com
- 11Apple interview questions
glassdoor.com
- 12Apple Go Concurrency Interview Questions
teamblind.com
- 13
- 14Google L6 phone screen
teamblind.com
- 15Why is my Anthropic offer barely above my Google L5?
teamblind.com
- 16Apple comp WTF
teamblind.com
- 17Software Engineer Interview, Apple, Oct 5, 2026
glassdoor.com
- 18Software Engineer III Interview, Google, Oct 4, 2026
glassdoor.com
- 19
- 20Software Engineer Interview, Google, Oct 4, 2026
glassdoor.com
- 21SDE Intern Interview, Apple, Oct 3, 2026
glassdoor.com
- 22Hardware Engineer Interview, Apple, Oct 3, 2026
glassdoor.com
- 23Software Engineer Interview, Microsoft, Oct 1, 2026
glassdoor.com
- 24Software Engineer Interview, Microsoft, Sep 30, 2026
glassdoor.com
- 25Software Engineer Interview, Microsoft, Sep 29, 2026
glassdoor.com
- 26Amazon Software Engineer Salaries
levels.fyi
- 27Stripe Software Engineer Salaries
levels.fyi
- 28
- 29
- 30Need advice about leaving current job
reddit.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.