Sep. 21–28 interview digest: AI-assisted rounds at Microsoft, Google and Amazon, and an offer revoked mid-H-1B transfer
This week's reports turn a one-day Microsoft onsite, four companies' AI-tooled rounds, an offer revoked during a visa transfer, and a $20K slide in Stripe's median into drills and negotiation references.
The seven days ending Sep. 28, 2026 put the word AI on the interview schedule itself. Microsoft has a round its own invitation calls "AI assisted interview." Google has an "AI enabled" format. LinkedIn's staff loop lists a round named "Code with AI." And an Amazon candidate reported that the online assessment let them use an internal AI assistant to debug the problem. 1234 In the same week, from the other side of the table, an interviewer described how he catches candidates using AI mid-round. That post drew 149 likes, 110 comments and 7,508 views, the largest audience of anything interview-related in this window. 5
Glassdoor's report pages were readable this window, and they carry what Blind does not: an absolute posting date, the candidate's own outcome, and the platform's difficulty and experience labels. Blind supplied first-person accounts with relative age markers (
2d, 3d, 4d, 7d) plus a like, comment and view counter under each post, which is the only per-report engagement figure that exists this week. Reddit supplied verified scores and timestamps. Levels.fyi refreshed all four software-engineer pages on Sep. 28.Four limits hold across the issue. Five of the twelve dated Glassdoor reports describe loops run in August 2026 or earlier, so the posting date decides membership and the interview month is printed beside it. No source publishes a cross-platform vote total, so the two report tables below are ordered inside their own platform and the counters never combine. Glassdoor prints no engagement figure on an individual report, so its rows carry a date and an outcome instead. And nobody publishes the applicant volumes and completed-interview counts that a company-wide or round-by-round pass rate would require, so this issue reports individual outcomes together with the conditions that produced them.
The week in one read
- Microsoft ran a full onsite in a single day. A candidate who applied online and was submitted to the August hiring event sat three back-to-back virtual interviews of about 45 minutes each, in the order system design, low-level design, then data structures and algorithms. The result was no offer, and the candidate still rated the experience positive and the interview difficult. 6
- A Microsoft screen can be two problems in one hour. A report dated Sep. 23 describes an online round of roughly an hour holding two LeetCode questions, one easy and one medium: generic string manipulation first, then a question built on chessboards. No offer. 7
- A Microsoft offer was withdrawn mid-H-1B transfer. The poster accepted an offer in mid-July, started the transfer, and then received a generic email saying he was no longer being considered. The post reached 6,044 views and drew 113 comments, and Blind's generated summary of them, dated Sep. 25, 2026, has commenters pointing at H-1B processing delays and telling him to go back to the recruiter for a specific reason. 8
- An Apple Senior iOS loop turned on unfinished code. Across five rounds, the candidate reasoned well through an async image loader — concurrency, partial failures, cancellation, ordering, stale results — while the interviewer kept changing requirements, and then overbuilt a vending machine design before simplifying it, running out of time with syntax errors in the tests. The candidate asked the community how heavily a hiring manager weighs incomplete implementations against strong reasoning. 9
- Ten-plus years of experience, one offer, $150K. A poster who had been laid off four times traced total compensation from $875K in 2021 to $550K in 2024, $320K in 2025, $260K in 2026 and zero now, then described roughly 80 interview processes this year producing a single offer of $120K base plus a $30K target bonus with no equity. Commenters split between accepting it and holding out. 10
- A Google candidate passed the phone screen and lost the loop anyway. The coding question was a medium solved in optimal time and the behavioral round went well; the recruiter confirmed a pass the next day. A couple of weeks later the candidate received an email citing changed business needs, and asked whether the promised follow-up ever comes. 11
Evidence at a glance
Loading stats card…
The twelve Glassdoor reports break down as three each at Microsoft, Amazon and Google, two at Apple, and one at Meta. All twelve were posted between Sep. 21 and Sep. 28, 2026. The twelve Blind posts are the ones whose age marker and counters were read off the page this window.
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
7d 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 |
|---|---|---|---|---|
| Employer not disclosed | 4d | Interviewer-side account of live coding rounds: the poster waits for a candidate to finish a polished solution, then asks them to explain a line and to change one constraint. | None reported. The poster describes failing candidates he judges to be reading an AI-generated answer. | 149 likes, 110 comments, 7,508 views. The poster's tag is Meta; the post sits in Blind's Tech Industry channel and names no hiring company. 5 |
| Employer not disclosed | 3d | Roughly 80 interview processes in 2026 after four layoffs; the single resulting offer was $120K base plus a $30K target bonus with no equity. | Offer held, decision pending. | 43 likes, 142 comments, about 10,000 views. Previous compensation is the poster's own account; the employer making the offer is unnamed. 10 |
| Microsoft | 7d | Offer accepted in mid-July, H-1B transfer in progress, then a generic email withdrawing consideration. | Offer revoked. | 21 likes, 113 comments, 6,044 views. Blind's generated summary draws on 111 comments and is dated Sep. 25, 2026. The poster is tagged Amazon and names no level. 8 |
| Microsoft | 4d | L62 verbal offer received; the poster is looking for a hiring manager because the recruiter can only circulate his profile where there is interest. | Offer in hand, team matching under way. | 1 like, 11 comments, 666 views. No compensation figures and no interview rounds described. 12 |
| Apple | 2d | Senior iOS loop, five rounds. Round content: an async image loader with requirements changed mid-round, and a vending machine design. | Not reported; the candidate is awaiting a verdict. | 3 likes, 10 comments, 618 views. Blind's generated summary of 10 comments is dated Sep. 27, 2026 and records commenters noting Apple's tendency to down-level. 9 |
7d | Phone screen passed, then the loop was canceled two weeks later with a message about changed business needs. | Loop canceled before the onsite. | 6 likes, 2 comments, 444 views. The poster's first big-tech loop; the team and the level are unnamed. 11 | |
3d | Staff loop of five rounds: Coding, Code with AI, System Design, Craft Skills, Managerial, scheduled for the following week. | Not reported; the poster was asking what to expect. | 2 likes, 11 comments, 386 views. Blind's summary of 10 comments, dated Sep. 27, 2026, records a commenter who was asked to implement a JSON parser on a tight clock and others who expect LeetCode hard problems. Comment-level content, not the poster's own account. 3 | |
| Meta | 4d | Loop in progress: one behavioral, one system design and one coding round done, with one system design and one coding round remaining. | Not reported; the candidate had completed part two of the coding round. | 0 likes, 7 comments, 270 views. Round structure only; no questions disclosed. 13 |
| Meta | 2d | A technical screen that produced no reply for 10 days, with one follow-up email unanswered. | Pending, with no response. | Age and channel read from the post page; the poster is tagged Qualcomm and no round content or counter was captured. 14 |
| Nvidia | 3d | CUDA core team panel across four interviewers: the hiring manager on a project deep dive, a medium LeetCode problem and CUDA experience; two engineers on data structures and C++ concepts; the final interviewer on a résumé deep dive and behavioral questions. | Not reported; the poster rated one round bad for missing the optimal solution. | 0 likes, 12 comments, 199 views. Blind's summary of 10 comments, dated Sep. 26, 2026, notes that the hiring manager was a senior director and that commenters think a manager's backing can outweigh one weak round. 15 |
3d | L4 onsite finished; the recruiter moved the candidate to a formal review taking one to two weeks, described in the post as possibly the hiring committee, ahead of team matching. | Pending review. | 0 likes, 9 comments, 180 views. The poster was asking Googlers whether a direct route to the committee implies stronger feedback ratings. 16 | |
| Wise | 3d | Senior pair-programming round asking for a circuit breaker. The pre-interview instructions named data structures and algorithms; the poster coded a single-threaded solution first and planned to add locks. | The poster expects a rejection after the interviewer stopped him mid-execution and called the solution good. | 0 likes, 2 comments, 55 views. A single account with the round described in the poster's own words. 17 |
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.
| Company | Report date | Round, task, or process | Reported outcome | Ratings attached |
|---|---|---|---|---|
| Sep 28, 2026 | Software engineering internship for 2027, North Plainfield, NJ. One technical round, LeetCode style, run on a Google-Docs-like platform. The one question listed is trees and backtracking with DFS. | No offer | Difficult; positive experience 18 | |
| Amazon | Sep 27, 2026 · interviewed Jun 2026 | Online assessment in four parts, three of them coding, the fourth long and built on work scenarios and simulations. The whole process ran three months, and the answer took two of them. | No offer | Difficult; neutral experience 19 |
| Sep 27, 2026 | Forward-deployed engineer internship. Recruiter screen asking the candidate to state their interest and to answer what Google is now that ChatGPT, Claude and DeepSeek exist. Further rounds may follow. | Accepted offer | Easy; positive experience 20 | |
| Microsoft | Sep 26, 2026 · interviewed Aug 2026 | Virtual onsite, three back-to-back interviews of about 45 minutes on one day, through the August hiring event: system design, then low-level design, then data structures and algorithms. | No offer | Difficult; positive experience 6 |
| Sep 26, 2026 | Senior data scientist. Basic HR phone screen on salary expectations and background, a technical video call on Python and SQL, then a virtual onsite with the hiring manager and two team members on past projects, machine learning concepts and cultural fit. About three weeks end to end. | No offer | Average difficulty; positive experience 21 | |
| Amazon | Sep 25, 2026 · interviewed Aug 2026 | Hiring panel described as kind and helpful; the poster's first interview, which he rated surprisingly relaxed. | No offer | Difficult; positive experience 22 |
| Amazon | Sep 24, 2026 · interviewed Jun 2026 | Online assessment that let the candidate use an internal AI assistant to debug the problem, mixing an architectural problem with data structures and algorithms. Process took one week. | No offer | Difficult; positive experience 23 |
| Microsoft | Sep 23, 2026 | Two LeetCode questions in about an hour on an online platform: generic string manipulation, then a problem built on chessboards. | No offer | Difficult; neutral experience 7 |
| Meta | Sep 23, 2026 | First interview completed with positive feedback, then an invitation to a live coding test. The poster has not finished the stages. | Still in process | Easy; positive experience 24 |
| Microsoft | Sep 22, 2026 · interviewed Nov 2024 | Three rounds in Hyderabad: data structures, computer science basics and databases; low-level design, where a problem statement arrives and the candidate defines classes, functions and design patterns; and high-level design built from functional and non-functional requirements. | Accepted offer | Average difficulty; positive experience 25 |
| Apple | Sep 22, 2026 | Austin, TX. A casual HR phone call covering background and years of experience as a baseline check before any technical stage. | No offer | Easy; neutral experience 26 |
| Apple | Sep 21, 2026 · interviewed Jul 2025 | Austin, TX. Recruiter call, technical screen with a team member, general screen with the hiring manager, then a culture call. The poster described poor communication about what each stage would contain. | Accepted offer | Average difficulty; negative experience 27 |
The question bank you can practice now
1. A three-round onsite in a single day
Microsoft's Seattle report lays out the order plainly: system design, low-level design, then data structures and algorithms, each about 45 minutes, back to back. 6 The Hyderabad report describes the same three disciplines from the inside: define classes, functions and design patterns against a given problem statement for the low-level round, then build a system from functional and non-functional requirements for the high-level one. 25
Three 45-minute rounds leave no room for setup. A drill that matches the clock:
- Pick one low-level prompt — a parking lot, an elevator bank, a rate limiter — and in 15 minutes write only the class and function signatures, then spend five minutes naming the design patterns you chose and why.
- Pick one high-level prompt and in 15 minutes write the functional requirements, the non-functional requirements, and one capacity number. State which component you would shard first and what you would monitor.
- Do both in the same sitting, alternating, so the switch between the two registers is itself rehearsed.
2. Two problems in one hour
The Sep. 23 Microsoft screen held an easy string-manipulation problem and a medium problem built on chessboards, inside roughly an hour. 7 Two problems in an hour means the first one has to close in twenty minutes including the explanation. Time five easy problems to a 20-minute clock each, then do a paired set back to back without a break.
Chessboard-shaped questions reward knowing two coordinate moves cold: the index arithmetic that maps a square to its row and column, and the two diagonal families that the row-plus-column and row-minus-column sums encode. That pair of facts collapses most board problems into a lookup.
3. Trees and backtracking on a blank editor
Google's internship report lists one question, trees and backtracking with DFS, asked on a platform the candidate described as Google-Docs-like. 18
An editor with no autocomplete, no bracket matching and no test runner removes every crutch a local IDE provides. Practice in a plain document for a week: write the recursive DFS, the visited set, and the backtracking undo by hand, then trace one example on paper before running anything. Five problems, no IDE.
4. Apple's Senior iOS round: an async image loader under moving requirements
The report is unusually specific about the two rounds that went wrong. In the first, the candidate reasoned correctly through concurrency, partial failures, cancellation, ordering and stale results, then the interviewer kept changing requirements and the final implementation never got rewritten. In the second, a vending machine design was overbuilt, then simplified, and the tests ran out of time with bracket errors. 9
Both failures are execution failures, and both are cheap to rehearse:
- Write an async image loader with a cache, cancellation and stale-result suppression in a single sitting, then change the requirement twice while writing and keep the code compiling.
- Under a running clock, write the tests for a second implementation you just wrote, and count how many syntax errors you make. Bracket errors under time pressure are a habit, and habits respond to repetition.
- Answer the level question before the loop: know which parts of your experience demonstrate concurrency and debugging at a senior level, because a five-round loop gives you four chances to show it and one weak round to lose it.
5. Rounds whose pre-instructions hide the real bar
Wise's senior pair-programming round asked for a circuit breaker. The pre-interview instructions mentioned data structures and algorithms; concurrency went unmentioned. The poster coded a single-threaded version first, intending to add locks, and the interviewer stopped him mid-execution to say the solution was good — which he read as the end of the assessment. 17
The lesson transfers to any senior round: ask which properties of the solution are being assessed before writing. A question that takes one sentence to ask — "do you want this correct under concurrency, or is single-threaded acceptable for now?" — converts a hidden expectation into a stated one. Prepare three such questions and use one at the start of every design-heavy round.
The format is changing: AI inside the round
Four companies now put an AI tool or an AI stage inside the interview itself, and the four reports describe four different shapes.
- Microsoft labels the round. A candidate heading into a loop found a round that its own schedule called "AI assisted interview" and asked what it measures, how much to delegate to the model, and what to do unaided. The post drew no comments, so no account of the round's content surfaced this window. 1
- Google labels the format. An engineer between jobs described an "AI enabled interview" and asked the same three questions — what he is measured against, how much to delegate, and what to do himself. That post also drew no replies. 2
- LinkedIn schedules it as one of five rounds. The staff loop contains a round named "Code with AI" between Coding and System Design, and a commenter under the post said the round involves coding on a laptop. 3
- Amazon permits it in the assessment. The candidate reporting the Sep. 24 software developer assessment wrote that the online stage allowed an internal AI assistant to help debug the problem, alongside an architectural problem mixed with data structures and algorithms. 23
A format label is not a preparation guide. Three of these four reports are questions asking the community, and the fourth describes the tool's availability rather than the questions asked, so none of them contains what the round actually set. What a candidate can do now is prepare the part that does not depend on the label: the ability to justify a line of code, to modify a solution under a changed constraint, and to say which portions you wrote and which the model produced.
That last skill showed up from the interviewer's side the same week, in a post that reached 7,508 views. The poster lets a candidate finish, then says, in his words, "great, now walk me through why you chose this line," and then changes one constraint. He reports that one candidate asked to start over three times and another claimed an unstable connection and left. 5 Under that post, the two most-liked replies went opposite ways: one commenter with 62 likes called the enjoyment itself concerning, and an Amazon commenter with 112 likes wrote that a candidate who can explain their code gives the interviewer a strong signal.
After the loop: when the offer moves
Three of this week's reports describe a decision reversing itself after the candidate had already acted.
- The offer that was withdrawn during a visa transfer. The poster accepted in mid-July, was mid-transfer when the withdrawal email arrived, and asked the community for openings in their own team at the same company. Blind's generated summary of the 111 comments, dated Sep. 25, 2026, has commenters pointing to H-1B approval delays following recent executive orders and lawsuits, and advising the poster to press the recruiter for a reason rather than a template. 8 For anyone in a transfer, the offer's conditions belong in writing before you resign anything, and the process belongs in email, because a generic withdrawal carries no reason a candidate can act on.
- The loop canceled after a passed phone screen. Google confirmed the pass, then sent "business needs changed" a couple of weeks later. 11 A phone-screen pass is a gate, and the headcount behind it can move. Keep other loops alive until a written offer exists.
- The screen that goes quiet. A candidate elsewhere wrote that 10 days had passed since a Meta technical screen with no update and no reply to a follow-up. 14
The same fragility runs through this week's Reddit threads. One poster counted four separate finalist-round rejections for senior roles, each after a loop of five to eight rounds, the most recent ending in an in-person whiteboard onsite where he judged his coding round strong and his system design round 85 percent right, followed by two generic rejection emails. 28 That post carries 342 points, 149 comments and an upvote ratio of 0.97.
Compensation: what changed this week
Reported offers and decisions
| Company | Situation | Numbers | Status |
|---|---|---|---|
| Amazon | A 28-year-old remote engineer in New York City with 6 years of experience holds a $200K total-compensation role with good balance and job security, and received a Seattle offer from a team that requires the move. The poster says he was down-leveled to mid-level and that almost every other application stalled at the résumé stage. | $200K remote versus about $280K in Seattle | Open, with 279 comments on it 29 |
| Unnamed employer | Ten-plus years of experience, four layoffs, about 80 processes this year, one offer: $120K base plus a $30K target bonus, no equity. | $150K all in | Open; the poster asked whether to take it or hold out 10 |
| Microsoft | A verbal offer at L62, with the recruiter circulating the profile for team interest before it becomes concrete. | Level only; no components disclosed | Verbal, team matching under way 12 |
Fresh aggregate anchors, refreshed Sep. 28, 2026
Levels.fyi refreshed all four software-engineer pages on Sep. 28, 2026. These are medians of self-reported submissions that the platform reviews, and they are the market reference a candidate brings into a counter. They are not signed offers.
| Company | Level | Total | Base | Stock | Bonus |
|---|---|---|---|---|---|
| L3 / SWE II (Entry) | $208K | $157K | $37.3K | $14.1K | |
| L4 / SWE III | $305K | $191K | $88.0K | $25.6K | |
| L5 / Senior SWE | $445K | $234K | $174K | $36.7K | |
| L6 / Staff SWE | $697K | $265K | $385K | $47.2K | |
| Meta | E3 (Entry) | $179K | $144K | $29.5K | $5.3K |
| Meta | E4 | $299K | $185K | $95.4K | $18.3K |
| Meta | E5 | $444K | $225K | $191K | $28.1K |
| Meta | E6 | $735K | $275K | $419K | $41.3K |
| Amazon | L4 / SDE I | $188K | $141K | $35.9K | $11.2K |
| Amazon | L5 / SDE II | $275K | $181K | $91.0K | $2.8K |
| Amazon | L6 / SDE III | $390K | $213K | $174K | $2.7K |
| Amazon | L7 / Principal SDE | $662K | $266K | $391K | $5.0K |
| Stripe | L1 (Entry) | $212K | $146K | $45.1K | $20.5K |
| Stripe | L2 | $287K | $188K | $85.2K | $14.4K |
| Stripe | L3 | $448K | $219K | $202K | $27.1K |
| Stripe | L4 / Staff Engineer | $662K | $274K | $330K | $58.4K |
Loading chart…
The company medians after the Sep. 28 refresh sit at $301K for Google, $455K for Meta, $272K for Amazon and $260K for Stripe. 30313233
Against the same four pages read a week earlier, three companies moved up and Stripe moved down. Meta's median went to $455K from $440K, Google's to $301K from $293K, and Amazon's to $272K from $266K, while Stripe's fell to $260K from $280K. At level granularity, Google's L6 rose to $697K from $654K and its L4 to $305K from $294K; Meta's E6 rose to $735K from $714K and its E5 to $444K from $430K; Amazon's L7 rose to $662K from $641K; and Stripe's L3 fell to $448K from $472K while its L4 held at $662K. Every figure is a platform refresh of an aggregate, so read the direction as a market reading and the number as one submission pool's median on one day. 30313233
What those totals pay in the years that matter depends on the vesting schedule behind them. Google's pages list front-loaded four-year schedules of 38/32/20/10 or 33/33/22/12 depending on share count. Meta vests quarterly over four years, or over five for some grants. Amazon back-loads at 5/15/40/40, which places 80 percent of the grant in years three and four. Stripe lists one-year, two-year and four-year schedules with a refresh available after nine months. These schedules come from the company pages and describe page-level context rather than the terms of any single offer. 30313233
What Glassdoor reports about outcomes
The number a candidate usually wants first — a pass rate — has no source. Glassdoor collects per-report outcomes and per-company experience ratings instead, and those refreshed this week.
| Company | Interviews on the page | Difficulty (out of 5) | Positive experience | Average days to hire |
|---|---|---|---|---|
| 26,144 | 3.36 | 63.6% | 38 | |
| Amazon | 5,259 | Not published on the page read | Not published on the page read | Not published on the page read |
| Microsoft | 1,720 | 3.2 | 64% | 29 |
| Meta | 2,281 | 3.3 | 56% | 27 |
| Apple | 701 | 3.2 | 56% | 23 |
Each figure is a rolling average across every report the platform holds rather than a measurement of this week's candidates, and the pages refresh at different times: Google's on Sep. 28, Amazon's on Sep. 27, Microsoft's on Sep. 26, Meta's on Sep. 23 and Apple's on Sep. 22. Across the twelve dated reports above, three ended in an accepted offer, eight in no offer, and one is still in process. That split describes twelve self-selected candidates rather than a company's hiring bar.
Company read-through
| Company | Current-week signal | What a candidate can use | Evidence boundary |
|---|---|---|---|
| Microsoft | The busiest company in the window: a three-round one-day onsite, a two-problem screen, a three-round loop ending in an accepted offer in Hyderabad, and an accepted offer withdrawn during an H-1B transfer. 68 | Rehearse the three disciplines in one sitting, and treat a verbal offer as provisional until the written terms and any transfer steps are confirmed. | Four accounts from Seattle, Hyderabad and an unnamed location; the two reports that describe the same three-discipline loop ran in August 2026 and November 2024. |
| One pass that led nowhere, one candidate waiting on a formal review, and statements of interest rather than round content: the two most-read format posts ask what "AI enabled" means and drew no replies. 211 | Keep loops running until a written offer exists, and practise coding in a bare document editor because at least one Google screen ran on a Docs-like platform. | Of the three Google reports, the accepted offer is an internship and the other two ended without one. | |
| Amazon | An assessment that permitted an internal AI assistant for debugging, a four-part online exam inside a three-month process, and one offer decision built on a down-level. 2329 | Price the Seattle move against the $80K delta and the stated down-level, and expect the assessment to mix architecture with algorithms. | Three self-reported assessments from Tel Aviv, Johannesburg and Cambridge, two of them run in June 2026, plus one offer decision at 6 years of experience. |
| Apple | Two Austin reports with opposite outcomes and a five-round Senior iOS loop hanging on unfinished code. 927 | Practise finishing implementations under changing requirements, and ask the recruiter to describe each stage before the loop starts. | Two reports from one city and one five-round account with no verdict yet. |
| Meta | A loop structure of two system design and two coding rounds, one report at the live-coding stage, and a screen that went quiet for 10 days. 131424 | Expect the design load to equal the coding load at senior levels, and keep the loop moving while waiting on a response. | Three unresolved accounts; no Meta outcome arrived this window. |
| Nvidia | A four-interviewer panel for a CUDA core team, with the hiring manager at senior-director level running the project deep dive. 15 | Prepare the domain depth separately from the algorithms: the panel split cleanly between CUDA, C++, and general data structures. | A single post with 199 views and no outcome. |
| LinkedIn and Wise | LinkedIn schedules five rounds including "Code with AI"; Wise asks for a circuit breaker in a pair-programming round whose instructions mentioned only data structures and algorithms. 317 | Ask which properties of a solution are being assessed before writing, and treat the round list as the syllabus. | Both are pre-outcome accounts with small counters. |
Before your next interview or negotiation
- Rehearse the three disciplines in one sitting. System design, low-level design and algorithms came back to back within one day at Microsoft. Fifteen minutes on signatures and patterns, fifteen on requirements and one capacity number, alternating.
- Time two problems to one hour. An easy problem plus a medium one inside 60 minutes means the first closes in 20, explanation included. Practise paired sets without a break.
- Code in a blank editor once a week. At least one Google screen ran on a Docs-like platform with no autocomplete and no test runner.
- Ask which properties are being assessed before you write. A senior pair-programming round asked for a circuit breaker while the instructions named only data structures and algorithms.
- Prepare to justify, not just to produce. The most-read hiring-process post this week is an interviewer asking candidates to explain a line and then change a constraint. Be ready to describe which parts of a solution you wrote and why.
- Separate the offer from the decision. Ten-plus years of experience produced one $150K offer this week, and a Microsoft offer was withdrawn mid-transfer. Keep interviews running, and get conditions in writing before you resign anything.
- Quote the refreshed medians when a recruiter names a number. Google $301K, Meta $455K, Amazon $272K and Stripe $260K as of Sep. 28, with Meta's E6 at $735K and Google's L6 at $697K. Use them as the market reference for the conversation rather than as a target.
References
- 1Microsoft AI assisted interview format
teamblind.com
- 2Google AI enabled interview
teamblind.com
- 3LinkedIn staff engineer loop round
teamblind.com
- 4Amazon Software ENgineer interview questions
glassdoor.com
- 5
- 6Software Engineer Interview, Microsoft, Sep 26, 2026
glassdoor.com
- 7Software Engineer Interview, Microsoft, Sep 23, 2026
glassdoor.com
- 8
- 9Finished Apple Loop. Need help assessing my chances!
teamblind.com
- 10Accept 150k offer or stay homeless?
teamblind.com
- 11
- 12Microsoft verbal offer
teamblind.com
- 13Meta interview
teamblind.com
- 14Meta tech screen still no update
teamblind.com
- 15Hiring decision after panel interview at Nvidia
teamblind.com
- 16
- 17Wise Pair Programming, need to know concurrency?
teamblind.com
- 18Software Engineer Interview, Google, Sep 28, 2026
glassdoor.com
- 19Software Engineer Interview, Amazon, Sep 27, 2026
glassdoor.com
- 20FDE Intern Interview, Google, Sep 27, 2026
glassdoor.com
- 21
- 22Software Engineer Interview, Amazon, Sep 25, 2026
glassdoor.com
- 23Software Developer Interview, Amazon, Sep 24, 2026
glassdoor.com
- 24Software Engineer Interview, Meta, Sep 23, 2026
glassdoor.com
- 25Software Engineer Interview, Microsoft, Sep 22, 2026
glassdoor.com
- 26Software Engineer Interview, Apple, Sep 22, 2026
glassdoor.com
- 27Software Engineer Interview, Apple, Sep 21, 2026
glassdoor.com
- 28Senior Engineer Job Market
reddit.com
- 29
- 30Google Software Engineer Salaries
levels.fyi
- 31Meta Software Engineer Salaries
levels.fyi
- 32Amazon Software Engineer Salaries
levels.fyi
- 33Stripe Software Engineer Salaries
levels.fyi
- 34Google interview questions
glassdoor.com
- 35Microsoft Software Engineer interview questions
glassdoor.com
- 36Meta Software Engineer interview questions
glassdoor.com
- 37Apple Software Engineer interview questions
glassdoor.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.