Oct. 5–12 interview digest: LinkedIn's AI-assisted coding round, and a 12-year veteran rated L5
This week's reports name LinkedIn's AI-assisted coding round inside a staff loop, price a Google L6 offer at $650K, and show a 12-year veteran rated a level below the role he applied for.
The seven days ending Oct. 12, 2026 put a name on the AI-assisted round. A candidate who finished the full LinkedIn Staff Software Engineer process listed five rounds, and one of them was the AI Assisted Coding Round. 1
He missed a defect in the AI-generated code, caught and fixed it after a hint, and said the follow-up went well. 1 Two days later a Senior SWE heading into the same round asked how much of the implementation may come from the model and what interviewers weigh. 2
Two reports this week turn on the level rather than the years. A candidate with 12 years of experience and a P50 title at Atlassian cleared Google's L6 phone screen, and both interviewers rated him L5 while the recruiter called the feedback positive and moved him to onsites. 3
In the same window, a candidate with 11 years of experience posted a Google L6 offer worth $650K a year averaged across four years, after six months of preparation. 4
Glassdoor's report pages were readable this window for Google, Amazon, Microsoft and Apple, and Meta's software-engineer page carried no report dated inside the window. 56789 Blind supplied first-person accounts with relative age markers and a like, comment and view counter under each post, read off page screenshots. Reddit's current threads were HackerRank's public break with LeetCode and a 2,126-point study thread on AI productivity, so Reddit appears here through its discussion. 1011
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 years earlier, and the interview month is printed beside the date where the report gives one. No source publishes the applicant volumes and completed-interview counts a company-wide pass rate would need, so this issue reports individual outcomes with the conditions that produced them.
The week in one read
- A 12-year veteran passed a Google L6 phone screen and was rated L5. Both interviewers marked the candidate a level below the role he applied for, even as the recruiter called the feedback positive and scheduled onsites. 3
- A Google L6 offer landed at $650K. The poster reports 11 years of experience and six months of preparation, and the offer averages $650K a year across four years. 4
- A fresh Anthropic offer came in under one the same candidate declined in 2023. The 2023 offer was about $700K; the new one carries a $300K base and roughly $655K in total, at 7 years of experience. 12
- One company pushed an interview-cheating catch past a rejection. A poster describes a friend caught using AI inside a coding round and a company that flagged the finding formally and discussed notifying other employers. 13
- Apple hired a web engineer off a round that re-implemented an existing library function. Two technicals and one behavioral, all in JavaScript, with the hiring manager saying the offer was effectively settled before the interview ended. 14
Evidence at a glance
Loading stats card…
The eight dated Glassdoor reports are two at Google, two at Apple, three at Amazon and one at Microsoft, all posted between Oct. 5 and Oct. 12. Company aggregates printed on those pages: Google lists 26,214 interviews, a 3.36 difficulty out of 5, 63.7% positive and 38 days to hire 5; Amazon lists 66,955 interviews, 3.09, 60% positive and 27 days 6; Microsoft lists 18,117 interviews, 3.13, 67% positive and 30 days 7; Apple lists 15,181 interviews, 3.05, 64.4% positive and 31 days 8. Meta's software-engineer page shows 2,717 interviews and was last updated Oct. 5. 9
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. Counters arrived on different days, so a
5d post has had longer to accumulate views than a 2d one. The counter row is the only engagement measure Blind publishes, and it is not comparable with a Reddit score.| Company or scope | Freshness shown | Round, task, or process | Reported outcome | Engagement and evidence boundary |
|---|---|---|---|---|
5d | An L6 offer after 11 years of experience and six months of preparation: system design from HelloInterview, and LeetCode limited to the top 100 plus the last 12 months of Google-tagged questions. | Accepted offer; new total compensation $650K averaged across four years. | 127 likes, 79 comments, 11K views. The poster is tagged Oracle. The highest-engagement report of the window. 4 | |
| Anthropic | 5d | An offer comparison across three years: an Anthropic offer declined in 2023 at roughly $700K, against a new Anthropic offer with a $300K base and about $655K total. | New offer in hand; the poster reports 7 years of experience and a current $615K total, and calls the lower number a disappointment. | 33 likes, 80 comments, 11K views. The poster is tagged Meta. Blind's AI summary of 11 comments is dated Oct. 8, 2026. 12 |
| Unnamed company | 4d | A secondhand account rather than a first-person one: a friend used AI assistance inside a coding round, was caught, and the company flagged the finding formally and discussed telling other employers. | Rejection plus a formal integrity flag; the poster asks where the boundary sits. | 26 likes, 82 comments, 6,753 views. The poster is tagged Rakuten and names no company. Higher engagement than any first-person loop report except the two above. 13 |
| Several companies | 4d | A staff-plus candidate took onsite calls at multiple companies, ran four mock interviews with FAANG-level interviewers, and applied the feedback in the real rounds. | No conversions; the poster asks what to change about their system design. | 16 likes, 46 comments, 3,408 views. The poster is tagged Uber. Blind's AI summary of 12 comments is dated Oct. 10, 2026 and names the trade-off habit as the recurring note. 15 |
| Roblox | 4d | A seven-round process. In one coding round the interviewer arrived without an introduction, rolled her eyes at suggested approaches, sighed, tapped the desk, and left no time for questions. | Reached the end of seven rounds; the poster asks whether untrained solo interviewers run loops there. | 10 likes, 50 comments, 2,966 views. The poster is tagged Unity. 16 |
| Anthropic | 3d | An onsite the poster says they have run several times over recent years. One interviewer told him his system design was "pissing away capacity and money"; in another round the interviewer rocked in a chair while asking questions. | No outcome stated; the poster describes the culture as the deterrent. | 19 likes, 18 comments, 2,761 views. The poster is tagged Airbnb. 17 |
| Shopify | 3d | Five virtual rounds for a Senior ML Engineer role: pair programming on a refund-decision library with requirements added in stages; ML modeling on a Shop recommendation feed; ML systems design on latency, scale, feature freshness and cold start; a 40-minute project deep dive; and a life-story conversation. AI tools were allowed in the coding round. | Loop completed; the poster shares the shape and offers to answer questions. | 0 likes, 0 comments, 46 views. The poster is tagged Maven Clinic. The lowest-reach report here, included because it is the window's most detailed first-person description of an ML loop. 18 |
2d | Five rounds for a Staff Software Engineer role: algorithms on a graph, where a follow-up coding mistake ran out the clock; system design that ran out of time before the trade-offs; the AI Assisted Coding Round, where a missed defect was caught after a hint; a craftsmanship round; and a host-leader round on a project and a large-scale design. | No outcome yet; the poster asks whether weaker rounds produce a rejection or a down-level. | 2 likes, 8 comments, 547 views, with a 20-participant poll. The poster is tagged ServiceNow and reports 12 years of experience and a current total of 85 lakh rupees a year, about $87K. 1 | |
5d | An L6 phone screen: a coding round and a Googleyness round. Both interviewers rated the candidate L5 even though the recruiter reported positive feedback and moved him on to onsites. | Advanced to onsites at L5. | 1 like, 3 comments, 112 views. The poster is tagged Atlassian and reports 12 years of experience and a P50 title there. 3 | |
28m | A question rather than a report, posted ahead of a Senior SWE AI-assisted coding round at LinkedIn: how hard the problems run, how much of the implementation may come from the model, and which uses of AI beyond code generation interviewers encourage. | No report yet; the thread carried 2 comments when read, so it supports no judgment about the round. | 0 likes, 2 comments, 15 views. The poster is tagged Walmart. Four hours old at the time of reading. 2 |
On Glassdoor, ordered by report date
Glassdoor publishes the date the report was posted, the candidate's own outcome and ratings, and no per-report engagement figure. Where a report gives the month it was interviewed, that month is printed beside the posting date, because the loop itself may have run years before the report appeared.
| Company | Report date | Round, task, or process | Reported outcome | Ratings attached |
|---|---|---|---|---|
| Oct 12, 2026 · interviewed Jun 2022 | Chicago, IL. Three one-hour live coding rounds after the initial screening, for a Software Engineer 1 role, with some non-technical questions mixed in. | No offer; the candidate did not reach the HR round. | Difficult; neutral experience 19 | |
| Oct 11, 2026 · interviewed Nov 2025 | Sunnyvale, CA. Two 45-minute LeetCode-style rounds built on graph backtracking and hashing, under strict time limits, with no behavioral round and time left at the end for the candidate's questions. | Accepted offer | Average; positive experience 20 | |
| Amazon | Oct 10, 2026 | Hyderabad. A recruiter reach-out, then an online take-home assessment, cleared after a week, followed by a call to the office. | No offer | Average; neutral experience 21 |
| Amazon | Oct 10, 2026 | Vancouver, BC. A 30-problem LeetCode plan laid out over 30 days and grouped by the patterns bar-raiser rounds hit most. | Accepted offer | Difficult; positive experience 22 |
| Apple | Oct 6, 2026 | New York, NY. A web engineering role in JavaScript: two technical rounds and one behavioral, the technical asking the candidate to re-implement an existing library function. | Accepted offer, told during the interview | Average; positive experience 14 |
| Apple | Oct 5, 2026 · interviewed May 2025 | Santa Clara, CA. A recruiter call covering the team and what it wanted, then a hiring-manager call about past projects and similar products the candidate had built. | No offer | Average; positive experience 23 |
| Amazon | Oct 9, 2026 | An HR call, a one-hour technical phone screen, then a virtual onsite of two hours of coding, one system design round and one hiring-manager round. The candidate also reported HR scheduling in the wrong time zone and changing times at short notice. | No offer | Average; positive experience 24 |
| Microsoft | Oct 11, 2026 | A single day of back-to-back technical rounds: a LeetCode question, system design, and a walkthrough of features the candidate had built in past roles. | Accepted offer | Average; neutral experience 25 |
The question bank you can practice now
1. Re-read the model's code before you defend it
The LinkedIn staff candidate's AI Assisted Coding Round turned on a defect in AI-generated code that he missed and then caught after a hint, with the follow-up going well. 1 HackerRank has now staked its product on that skill: its chief executive announced a shift from algorithmic questions to real work in repositories, from code correctness to "AI fluency and judgment", and from an IDE to an agentic environment, with a tool that asks candidates why they chose an approach, what they would change, and how they validated the AI-generated code. 26
A drill that matches both:
- Take a solved problem, generate a solution with a model, and plant one defect in it. Then find it under a ten-minute clock, saying out loud what you checked and what you ruled out.
- Answer the three questions the agentic interviewer asks: why this approach, what you would change, and how you validated the generated code. If you cannot answer the third in two sentences, you do not yet own the code.
2. Code that has to run, against a clock
Amazon's virtual onsite ran two hours of coding, one system design round and one hiring-manager round, and the candidate who cleared the Vancouver bar did it off a 30-day plan of one timed problem a day, grouped by the patterns bar-raiser rounds hit. 2224
- Set 30 to 40 minutes for one problem, write it end to end, and run it. Read the editorial only after the timer stops.
- In 60 minutes, sketch one design you would defend at bar-raiser depth: the components, the failure mode of each, and the metric that would tell you it broke.
3. The round that is specific to the role
Apple's New York round was built for one job: a web engineering role in JavaScript, asked to re-implement an existing library function, with nothing off-script. 14 The Santa Clara report two days earlier shows the other half of Apple's process, where a recruiter call and a hiring-manager conversation about past projects decide whether a loop happens at all. 23
- Pick one library you use weekly, and be ready to re-implement its core function from a blank file, explaining the interface you chose and the edge cases you handled.
4. Graphs, and the clock that ends the round
Two rounds this window came down to graphs against a timer. Google's intern round was two 45-minute questions on graph backtracking and hashing, with strict limits and no behavioral portion. 20 The LinkedIn staff candidate solved the first part of a graph problem, found the right approach for the follow-up, then made a coding mistake he could not fix before time ran out, which he counts among his weaker rounds. 1
- Drill the follow-up, not the first answer: solve one part, then write and test the extension under a fresh timer. The mistake that costs the round happens in the second half.
5. Staff-level system design is a trade-off habit
A staff-plus candidate took onsite calls at several companies, ran four mock interviews with FAANG-level interviewers, and converted none of them. 15
The comments under that post split along a line worth knowing before a staff loop. One commenter argues that a design built from remembered diagrams reads as memorized, while another answers that most engineers never touch the internals of a system at YouTube's scale and asks how anyone is supposed to answer without reciting what they read. 15
- Take one design you already know, and open with the constraint you are trading against: cost against latency, consistency against availability. Name the option you rejected and why, before the interviewer asks.
The level is set by the loop
The Google phone-screen report is the clearest case of a level moving independently of the candidate. The poster has 12 years of experience and a P50 title at Atlassian, reports that he demonstrated broad scope in the Googleyness round, and still had both interviewers rate him L5 on an L6 application. His own reading of that is worth noting, because he says he does not want to run the promotion cycle again at a company where his current team already carries six P50s. 3
That $650K figure is the number to hold next to a table, because it arrives averaged across four years, which is not how the platforms below report. 4 Levels.fyi refreshed all four software-engineer pages on Oct. 12, 2026, and its medians are a single-year figure per level. Read against them, a $650K offer for Google L6 lands below the platform's own L6 median.
| Company | Median total (US) | Median by level | Submissions | Benchmark refreshed |
|---|---|---|---|---|
| $284K | L3 $212K · L4 $307K · L5 $450K · L6 $730K | 23,427 | Oct 12, 2026 27 | |
| Meta | $482K | E3 $186K · E4 $300K · E5 $462K · E6 $768K | 13,878 | Oct 12, 2026 28 |
| Amazon | $270K | L4 $181K · L5 $275K · L6 $388K · L7 $620K | 41,642 | Oct 12, 2026 29 |
| Stripe | $260K | L1 $225K · L2 $284K · L3 $459K · L4 $621K | 1,525 | Oct 12, 2026 30 |
The medians hide the mix inside them. Levels.fyi prints each level's parts, and the weight shifts toward stock as the level rises.
| Company and level | Total | Base | Stock per year | Bonus |
|---|---|---|---|---|
| Google L5 | $450K | $234K | $182K | $33.6K 27 |
| Google L6 | $730K | $278K | $390K | $62.1K 27 |
| Meta E5 | $462K | $225K | $206K | $31.5K 28 |
| Amazon L6 | $388K | $215K | $170K | $2.9K 29 |
| Stripe L4 | $621K | $279K | $297K | $45.4K 30 |
Amazon's vesting runs 5%, 15%, 40% and 40% across four years, so an Amazon L6 total arrives later than the headline number suggests. 29 These are platform medians drawn from self-reported submissions, and they set an expectation rather than an offer. An Anthropic offer reported this window shows how far the two can drift: the same candidate was offered roughly $700K in 2023 and about $655K in 2026, with a $300K base, at 7 years of experience. Commenters under that post argue that more candidates now want the lab, which changes what it has to pay, and one suggests taking a Google offer instead. 12
After the loop
Three reports this window describe what the candidate cannot control once the interviews end.
An integrity catch can travel further than a rejection. A poster describes a friend who used AI assistance inside a coding round, was caught, and then found the company treating it as more than a failed loop: the finding was flagged formally, and the poster says the company discussed notifying others through back channels. The poster knows the cheating happened and still asks where the boundary sits between a rejection, a reapplication ban, and a mark that follows someone between employers. Under it, 82 comments argue the question on both sides. 13
A finished loop can end with no answer at all. The LinkedIn staff candidate completed all five rounds and still cannot tell whether his weaker rounds point to a rejection or to a down-level, which is why he put the question to a poll. Twenty people answered, and the two options were rejection and down-level. 1
And the interviewer decides how much of the loop is even about the work. The Roblox report describes a coding round run by an interviewer who gave no introduction, rolled her eyes at suggested approaches, and left before the candidate's questions, inside a process the poster counts at seven rounds. 16 The Anthropic report makes the same point from the other direction, describing an onsite where one interviewer called the candidate's system design "pissing away capacity and money". 17 Both posters say the room cost them performance, and neither has a process fix to offer beyond choosing a company whose interviewers are trained.
What the AI-assisted round asks you to say
LinkedIn has given the AI-assisted round a fixed place in a staff loop, and the two reports about it agree on the shape of the work. The candidate who ran it described catching a defect the model left behind. The candidate about to run it asked which uses of AI interviewers reward and where the line sits between editing the model's output and passing it along. 12
HackerRank's announcement says the same thing from the vendor's side: the questions it wants to ask are why you chose an approach, what you would change, and how you validated the generated code. 26 That is the part of the loop you can rehearse alone this week, on code the model wrote and you have to justify.
References
- 1
- 2
- 3Google L6 phone screen result
teamblind.com
- 4Received Google L6 offer
teamblind.com
- 5Google interview questions
glassdoor.com
- 6Amazon interview questions
glassdoor.com
- 7Microsoft interview questions
glassdoor.com
- 8Apple interview questions
glassdoor.com
- 9Meta Software Engineer interview questions
glassdoor.com
- 10
- 11
- 12
- 13
- 14Software Engineer Interview, Apple, Oct 6, 2026
glassdoor.com
- 15System Design for staff+
teamblind.com
- 16Bizarre experience interviewing at Roblox
teamblind.com
- 17Interview at Anthropic, terrible experience
teamblind.com
- 18
- 19Software Developer Interview, Google, Oct 12, 2026
glassdoor.com
- 20SWE Internship Interview, Google, Oct 11, 2026
glassdoor.com
- 21Software Engineer Interview, Amazon, Oct 10, 2026
glassdoor.com
- 22
- 23Software Engineer Interview, Apple, Oct 5, 2026
glassdoor.com
- 24Software Engineer Interview, Amazon, Oct 9, 2026
glassdoor.com
- 25Software Engineer Interview, Microsoft, Oct 11, 2026
glassdoor.com
- 26Sorry, we made you grind LeetCode
hackerrank.com
- 27Google Software Engineer Salaries
levels.fyi
- 28Meta Software Engineer Salaries
levels.fyi
- 29Amazon Software Engineer Salaries
levels.fyi
- 30Stripe Software Engineer Salaries
levels.fyi
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.