Interview evidence · 2026-09-22 snapshot

Review date: 2026-09-22. Dated evidence boundary: 2025-09-22 or later; preferred window: 2026-03-22 through 2026-09-22. The selected firsthand-report table below uses the preferred window. Reports outside six months but inside one year may be retained separately with that limitation; older publications do not support recent conclusions. This is a dated, qualitative review, not a representative survey. We searched Reddit, LeetCode Discuss, first-person posts, interview-preparation sites, and official employer guidance. Old reports, recycled “2026” titles, unverified dates, and unrelated search results were not used to establish recent trends.

What this evidence supports

Prepare both algorithms and practical implementation. The dated Amazon report below combines coding, low-level design, system design, and behavioral questions. The Google report describes graph reasoning and coding. Experienced-engineer discussions also describe API work, code review, and AI-assisted exercises. For senior/staff practice, add deeper follow-ups and evidence from real projects.

It does not support a precise “most asked” ranking, universal company loops, or exact pass thresholds. Authors self-select, outcomes are self-reported, multiple comments can describe the same experience, and location/team differences are substantial. We do not treat a poster's interpretation of interviewer feedback as the company's rubric.

Dated evidence ledger

Read the original source before using a company-specific detail. Dates below distinguish the event from the post; an unknown event date stays unknown. All sources were accessed 2026-09-22.

ID	Source and date basis	Level / location	Narrow observation	Curriculum consequence
R1	Amazon SDE II experience — interview dates explicitly July 2, 27, 29, 31 and August 4, 2026; outcome August 11	SDE II; location not established	Prefix-sum question, cache implementation, dependency scheduling, device-notification design, and behavioral follow-ups	Prefix sums, LRU, topological order, failure scenarios, implementation timeboxes
R2	Amazon experience and feedback request — indexed publication April 24, 2026; individual interview dates unspecified	SDE II; location not established	Graph/tree tasks, substring window, feature-level read receipts, and an idempotency discussion	Windows/BFS; clarify whether interviewer wants feature implementation or whole-system design
R3	US Google L4 offer report — indexed publication June 18, 2026; individual event dates unspecified	L4, US; do not relabel L4 as senior	Weighted graph problem, clarifying discussion, implementation and optimization under time pressure	Dijkstra, dry runs, communicating assumptions
R4	What gets asked in 2026? — indexed publication July 8, 2026	OP 5 YOE, US big tech; commenters vary	Respondents describe coding/design/behavioral alongside some practical API, PR review, and AI-enabled work	Keep both independent and practical/AI coding exercises
R5	Preparing for senior/staff roles — indexed publication August 25, 2026	15 YOE, frontend/full stack/product; SF Bay/NYC targets	Author reports deeper design probes, project-story difficulty, and both manual and agentic coding assessments	Staff ownership rubric; full-stack mutations; changing-constraint mocks
R6	2026 interview experience — indexed publication July 2, 2026	7 YOE; mentions Anthropic and Databricks	Author found practical correctness under time pressure demanding	Test edge cases and finish runnable code; do not adopt the author's universal “perfect or fail” claim
R7	How is the interview process? — indexed publication June 1, 2026	Mixed companies/levels	A respondent describes no-AI live coding plus an AI-allowed assignment	Practice both modes; verify actual rules
R8	Designing experienced-dev interviews — indexed publication April 24, 2026	Interviewer perspective; mid/senior	Discussion favors probing actual decisions and alternatives in past projects	Decision-story questions; this is interviewer opinion, not a company mandate
R9	First-person preparation reflection — published March 24, 2026	Company/level not established	Author emphasizes communicating existing experience	Supplementary perspective only; no frequency claim or named-company inference

R1–R3 are three named-company candidate narratives, not three random samples of the market. R4–R8 are discussion threads; do not count every reply as an independent verified interview. R9 is a personal reflection. Indexed publication dates provide coarse recency, not verification of interview dates or identity.

Current official guidance, separate from recent reports

Source	What it explicitly supports	Date limitation
Amazon SDE III preparation	Coding, system design, high/low-level design resources, behavioral evaluation; executable code and testing	Live guidance accessed September 22; publication date not established
Microsoft technical interviewing	Problem solving, design, runnable coding, testing, and role-related experience	Live guidance accessed September 22; publication date not established

These pages explain current published expectations. They are not counted as interviews in the six-month window. AWS, Python, ECMAScript, and Kafka primary documentation are used only for technical correctness and appear next to the relevant lessons.

How questions are asked and how to practice beyond the baseline

Reported shape	Baseline practice	Deeper follow-up	Level emphasis
Coding with added requirements	Correct code and explicit complexity	Change memory/ordering/streaming constraints; retest	Junior: core invariant; senior/staff: preserve correctness under change
Feature design inside an existing product	Model state, interfaces, and normal flow	Concurrent edits, offline users, out-of-order events	Junior: coherent feature; senior: failure behavior; staff: compatibility and ownership
Whole-system design	Requirements and complete data flow	Identify actual bottleneck, cost, recovery, and security boundary	Increasing depth and independence; no universal company mapping
Project deep dive	Explain your contribution and outcome	Rejected alternative, disagreement, evidence, and a decision you would change	Senior: ownership; staff: cross-team direction and influence
AI-assisted or repository task	Run, understand, modify, and test code	Catch generated mistakes; explain every critical boundary	All levels: own correctness; senior/staff: scope and system consequences

These “deeper” columns are our teaching recommendations, informed by the sources, not official scoring criteria.

Excluded or unresolved leads

Lead	Why not used as current evidence
Meta E5 offer / problem list	February 3, 2026 is within one year but outside the preferred six-month sample; not used to infer current Meta format
Meta interview with AI-enabled round	November 25, 2025 is within one year but outside the preferred sample; no underlying event recency established here
Stripe backend interview	Useful-looking account, but publication/event date was not exposed in accessible text; no current Stripe claim is based on it
Google L4 August lead	Indexed date and rendered relative age disagree; excluded from dated trend conclusions
Generic “2026” lists and old reports in related-post panels	Title/crawl date does not establish a recent interview
Meta official preparation page	Could not retrieve the page in this research pass; no contents inferred

Coverage gaps: no adequately verified recent junior-specific or named-company staff-loop sample; limited geographic coverage; no reliable Meta/Stripe/Microsoft recent question-frequency data. The junior and staff routes therefore use clearly marked practice standards and current official/general evidence, not invented company intelligence. This snapshot is useful but does not satisfy a statistical “what is most asked across top companies” claim.

The separate system-design prompt catalog review adds 15 commonly cataloged architecture prompt types for practice. Its live user-submitted entries do not expose individual interview dates, so they are not counted in this recent-evidence sample and do not close the recency or frequency gaps above.

Refresh procedure

Before relying on this for a new application, check the recruiter packet. When updating the research, retain the old snapshot and record: source URL, publication date, interview date, company/team/location, level as stated, round type, narrow observation, and confidence. Exclude out-of-window evidence from recent conclusions. Cross-check a pattern across independent accounts before calling it recurring; never fabricate a percentage without a sampling method and denominator.

Trace older research claims before using them

The original research notes are now connected to a claim-level ledger that distinguishes exact source, primary/secondary status, publication/event/access date, sample/role and curriculum inference. Unsupported statistics and consensus claims are explicitly withdrawn. This does not add named-company interviews to the sample above. Technical documentation and production reports cannot fill a missing recent lead-loop sample.

Run python docs/research/check_claims.py for local links and ledger structure; it does not verify source quality or event dates. Before using a company-specific assertion, read the exact source and ask which detail actually changes practice. A newly published repost of an older interview belongs to the older event; an undated live official guide has publication age unknown.

Sources and further reading · 15