FAANG interview digest, May 12–18: Google lets Gemini in the room, Meta prepares 8,000 more cuts
Google confirmed a pilot letting SWE candidates use Gemini during a coding round — and now grades AI fluency. Meta's 8,000-person layoff round lands around May 20. Plus: Amazon's actual question bank this week (GenAI deep-dive + heap/DP problems), Nvidia's ghosting-then-callback pattern, and a 10-company TC table for L4–L6.
FAANG interview digest, May 12–18: Google lets Gemini in the room, Meta prepares 8,000 more cuts, and what hiring actually looks like right now
This week the interview meta shifted in a concrete way. Google officially confirmed it's piloting a format where candidates can use Gemini during a coding round — not as a cheat shortcut, but as a deliberate signal that "AI fluency" is now on the scorecard. Meanwhile, Meta is days away from announcing roughly 8,000 more layoffs, Amazon has 30,000 corporate cuts on its books but is still staffing 11,000 interns, and Nvidia keeps ghosting candidates after brutal onsites. Here's everything that matters before your next interview.
Google's Gemini pilot: what actually changes
Google's VP of Recruiting Brian Ong confirmed to Business Insider that the company is running an active pilot for a new interview format, beginning in H2 2026 with rollouts already underway at Google Cloud and the Platforms & Devices unit 1.
The changes, broken down:
- Code comprehension round: Candidates at the junior-to-mid level may use Gemini during a round focused on reading, debugging, and optimising an existing codebase. Interviewers will grade "AI fluency" — prompt construction, output validation, and catching AI-generated mistakes.
- Googleyness round revamp: The behavioral component now includes a technical design discussion about a past project, not just values-based questions.
- Junior candidate swap: One traditional technical round is replaced with an open-ended engineering challenge.
The internal document describes the format as "human-led, AI-assisted," which is a direct acknowledgment of where Google's engineering culture has moved. The company said in April that 75% of its new internal code is now written by AI 2.
What this means if you're preparing: pure LeetCode grinding without AI workflow practice is no longer sufficient for Google. You need to be able to work alongside an AI assistant, catch its mistakes, and explain why you overrode or accepted a suggestion. The debugging and optimization reasoning matters more than grinding to a clean solution alone.
Hiring map: who's open, who's frozen
The picture is messier than last week's clear "Amazon and Google are hiring" story.
Amazon is running a split personality. The company confirmed up to 30,000 corporate job cuts (14,000 announced in October 2025 plus 16,000 additional) 3. At the same time, AWS CEO publicly dismissed AI job loss fears and said Amazon plans to hire 11,000 interns in 2026 4. Individual SDE-2 and SDE-1 roles are actively recruiting, but the recruiter communication problem is real: one candidate reported a 4-month process — OA in February, onsite rounds in March and April — with month-long gaps between each round and no response to LinkedIn messages 5.
Meta is the clearest freeze signal this week. WIRED confirmed that ~8,000 layoffs are expected to be announced around May 20, on top of roughly 25,000 cuts over the previous four years 6. In April, Meta forcibly transferred 1,000 of its top engineers into an Applied AI Engineering division inside Reality Labs; engineers who refused were laid off. The company also installed activity-tracking software (MCI) on US employee laptops to generate AI training data, with no opt-out option available — which triggered internal petitions and is now a morale signal that external candidates should factor in 7.
Google is visibly hiring, as evidenced by the interview pilot rollout. A reported Google L3 ML offer from May 2026 shows the current structure: a candidate with 3 years of experience at a semiconductor company (current comp $190k base + $40k bonus + $60k equity) received an L3 Google offer trending in May after a process that started in August 2025 and included two additional GPU programming rounds in March 2026 8. Candidates are getting caught on the L3-versus-L4 leveling question: Google's internal HRC process tends to level conservatively when headcount is tight.
Nvidia is interviewing but the post-onsite communication is broken. One candidate with 9 YOE reported an in-person onsite with "brutal rounds," weeks of ghosting while emailing the recruiter repeatedly, and then a sudden availability request a month later for the same role 9. TC for this candidate was $320k. The pattern — ghosting followed by unexpected outreach — has been seen enough at Nvidia that it's no longer surprising to the community.
Microsoft is showing negotiation-friendly signals. An Instagram post from the Blind account highlighted a candidate who got a promotion offer from Microsoft the week they handed in their resignation — a classic counteroffer pattern that suggests Microsoft's hiring bar for retention is currently lower than its external recruiting bar 10.
This week's question bank
Amazon SDE intern (two-round format, May 2026)
Both rounds followed the same structure: intro + GenAI project deep-dive + 2 DSA questions + 1–2 LP questions 11.
Round 1 questions:
- Gas station — greedy (LC medium equivalent)
- Boats to save people — two-pointer/greedy
- LP: Describe a conflict with a teammate and how you resolved it
- Project deep-dive: RAG architecture, why not just call ChatGPT directly, handling generic outputs, validating Gemini responses
Round 2 questions:
- Coupon heap problem: given a prices array and M coupons (each divides a price by powers of 2), minimize total cost → greedy + max-heap
- "House Robber with value transformation": if you take value x, you cannot take adjacent values x−1 or x+1; derive frequency accumulation, explain the recurrence on a specific test case before writing DP
- Follow-up: "Explain your complexity reasoning before you code it"
The GenAI project angle in both rounds is new and worth taking seriously. Amazon interviewers are now asking "why not just ChatGPT" as a genuine technical filter question — they want candidates who understand prompt chaining, output validation failure modes, and when to trust or reject a model's response.
LP pattern: the "conflict in a team" question appeared in round 1. Expect "Tell me about a time you had to push back on a technical decision" and "Describe a time a project failed — what did you do?" to appear across rounds.
AWS SDE-2 (phone screen)
The HM asked a rate-limiter + sliding window design question and asked the candidate to code it 12. One candidate who fumbled the coding portion (needed multiple nudges) still had the general SDE pool role "under consideration" after an auto-reject on the specific job ID. AWS appears to keep candidates in a common pool for team-matching even after a specific role rejection.
Google coding rounds (emerging from pilot)
Based on the documented pilot format 1:
- Code comprehension: read a provided codebase, identify bugs, propose and implement optimizations. Gemini is available — but interviewers grade whether you understood the code or just outsourced the reading.
- Open-ended engineering challenge (L3/junior track): no well-defined problem statement; you're expected to ask clarifying questions, scope it down, and reason about tradeoffs before touching code.
- Past-project technical design: pick a project you've built, explain the architecture, identify what you'd change now.
For the traditional rounds, this week's Reddit thread with Google DS & Algo questions from 2026 13 identified graph traversal, sliding window, and binary search as the most frequently appearing patterns.
Compensation data: what offers actually look like
| Company | Level | Median TC (2026) |
|---|---|---|
| Amazon | L4 (SDE I) | $191k |
| Amazon | L5 (SDE II) | $270k |
| Amazon | L6 (Senior SDE) | $392k |
| Amazon | L7 (Principal) | $627k |
| L5 (Senior SWE) | $370k avg ($346k–$579k range) | |
| All levels, 25th–90th pct | $195k–$430k | |
| Apple | ICT5 (Senior SWE) | $458k median ($360k–$570k+) |
| Apple | New grad SET (San Diego) | $140k base + $94k RSU/4yr + $10k sign-on |
| Meta | Median TC (2025 disclosed) | $388k (down from $417k in 2024) |
| Microsoft | SDE2 (L61–62) | $200k–$206k |
| Microsoft | Senior SDE (L63) | $233k |
| Microsoft | Principal SDE (L64) | $281k |
Sources: Levels.fyi, Blind, and direct candidate disclosures cited above.
A few things stand out. Amazon's backloaded vesting (5%/15%/40%/40% over 4 years) means L5 and below candidates should negotiate base heavily — you don't see the bulk of equity until years 3 and 4, and that's assuming no re-org or offer rescission. Meta's median TC dropped $29k year-over-year, which tracks with the reported compensation cuts to regular employees while the AI researcher salary band stretches to $100M/year. Apple ICT5 at $458k median is now above Meta's median.
The current negotiation window: Google's L3-versus-L4 leveling issue means candidates with 3+ years should push back explicitly on leveling before discussing numbers. The L3 floor and L4 floor differ by roughly $50–80k in TC at Google — that conversation is worth having before you see a number.
Patterns worth tracking going into next week
- AI fluency is now a testable skill at Google, and others are likely to follow. Being able to articulate when to accept, override, or discard an AI suggestion is a distinct competency from traditional coding ability — and you can practice it specifically.
- Meta's upcoming announcement on May 20 will be the biggest hiring signal of the next 30 days. If the cuts are concentrated in infrastructure and Reality Labs (as the April AI transfers suggest), that will free up experienced engineers into the market and push compensation benchmarks down in those roles.
- The 4-month Amazon loop is a feature, not a bug. Multiple candidates report month-long gaps between rounds. If you're in the Amazon funnel, keep the process alive via follow-ups every 2–3 weeks — the "under consideration for general pool" status is real and team matches can emerge from it.
- Nvidia's ghosting pattern: one month of silence after an onsite does not mean rejection. Maintain the recruiter relationship; the availability ping often comes out of nowhere.
References
- 1
- 2Google AI coding interview guide — Exponent
tryexponent.com
- 32026 Layoffs Tracker — WSJ
wsj.com
- 4
- 5
- 6
- 7Meta layoff cycle May 2026 — CV by JD
cv-by-jd.com
- 8
- 9
- 10
- 11
- 12AWS SDE-2 interview experience — Reddit
reddit.com
- 13
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.