Anthropic Software Engineer Interview Questions

Anthropic·Software Engineer·Online Assessment (OA)

Sep 2026

Applied for a Software Engineer role at Anthropic and completed a 90-minute CodeSignal progressive coding assessment, scoring 580/600. Got rejected anyway, which stung given the score. The whole thing left me wondering if they even looked at my resume before sending the assessment.

Build a small in-memory system incrementally, extending it across multiple parts while keeping all previously implemented behavior intact.

“The progressive structure is the whole trick.”

View Post

Anthropic·Software Engineer·Onsite - System Design / Architecture

Jul 2026

System design round at Anthropic for a software engineering role. The whole thing was basically one big question about web crawlers, but it kept expanding into distributed systems territory pretty fast.

Design a scalable web crawler that can discover and download pages across the public internet. Walk through the core architecture including the URL frontier, fetchers, parsers, and storage layer.
How would you extend this crawler to run across multiple machines with multithreading? Cover concurrency controls, per-host rate limiting, back-pressure, fault tolerance, and whether you're guaranteeing exactly-once or at-least-once processing.

“I started with the URL frontier and worked outward, which felt right.”

View Post

Anthropic·Software Engineer·Technical Phone Screen

Jul 2026

Coding round at Anthropic for a software engineer role. One algorithmic problem, felt pretty standard for the level but the edge cases tripped me up more than I expected.

Given a list of strings, find the maximum length of a concatenation of a subset of those strings such that every character in the result is unique across the whole thing.

“My first instinct was to just try all subsets and track which characters were used, and that's basically the right move since the input is capped at 16 strings.”

View Post

Anthropic·Software Engineer·Technical Phone Screen

Jul 2026

Anthropic had me implement a neural network from scratch, no libraries, no autograd, just math and code. It was the kind of question that sounds manageable until you're actually deriving backprop by hand under time pressure.

From first principles, implement a small neural network for binary classification with one hidden layer, including the forward pass, cross-entropy loss, and a full backward pass without any automatic differentiation library.
How would you verify that your manually derived gradients are correct using finite-difference gradient checking?
Discuss numerical stability considerations in this setting, such as the log-sum-exp trick, and how you'd apply them to cross-entropy loss.
What are the tradeoffs between different weight initialization strategies and activation function choices for a network like this?
How does batch size affect gradient variance during training, and what are the practical implications?

“This wrecked me a little.”

View Post

Anthropic·Software Engineer·Onsite - Coding / Algorithms

Jul 2026

Coding round at Anthropic for a software engineer role. One meaty algorithmic problem that looked manageable at first but had enough edge cases to keep you busy for the whole session, plus a follow-up that changed the problem just enough to break your first solution.

Implement a function that converts a sequence of timestamped stack samples into a list of start and end trace events. A start event fires when a function appears deeper in the stack than it was in the previous sample. An end event fires when a function drops off the stack, with inner frames ending before outer ones. Functions still on the stack at the last sample are assumed to still be running. You also need to handle identical consecutive stacks without emitting duplicate events, and treat recursive frames (same function name appearing at different depths) as distinct.
Follow-up: modify your solution so that a start event is only emitted for a frame that appears in at least N consecutive samples at the same position. You can configure N. Decide whether the start timestamp should come from the first sample where the frame appears or from the Nth. Make sure recursion still works correctly under this new rule.

“The recursion part tripped me up more than I expected.”

View Post

Anthropic·Software Engineer·Onsite - System Design / Architecture

Jul 2026

Anthropic system design round for a software engineer role. The whole thing was basically one long deep dive into a producer-consumer buffer, which sounds manageable until they keep pulling the thread on concurrency edge cases, fairness, and lock-free alternatives.

Design and implement a thread-safe, fixed-capacity producer-consumer buffer that supports multiple producers and consumers, with blocking put() and take() operations, FIFO ordering, configurable capacity, and a shutdown() method that unblocks waiting threads and rejects new puts.
How do you prevent race conditions, deadlocks, and lost wakeups in your buffer implementation?
Add timeout support to put() and take() so callers can specify a maximum wait duration.
Discuss the fairness and performance trade-offs in your design.
What are the time and space complexity of your buffer implementation?
Compare a lock-based design using a mutex and condition variables to a lock-free approach for this buffer.

“This felt like a warm-up until it wasn't.”

View Post

Anthropic·Software Engineer·Onsite - System Design / Architecture

Jul 2026

System design round at Anthropic for a software engineer role. One big open-ended question that basically ate the whole session, and they really did want you to go deep on everything.

Walk through a recent technical project you led or significantly contributed to. Cover the problem, constraints, architecture, key design decisions, trade-offs, data model, interfaces, scaling strategy, testing and validation, rollout plan, and how you measured success.
What was the hardest technical challenge in that project, how did you debug it, and what would you change in a second version?

“This is a monster of a question and I underestimated how much ground they actually wanted to cover.”

View Post

Anthropic·Software Engineer·Technical Phone Screen

Jul 2026

Anthropic software engineer interview with a meaty in-memory banking system design and implementation problem. The kind of question that looks manageable until you're 40 minutes in and realizing you forgot about transfer expiration cascading into merge edge cases.

Design and implement a full in-memory banking system supporting account creation, deposits, transfers with expiration, transfer acceptance, top outgoing accounts ranking, and account merging.

“This one is a beast.”

View Post

Anthropic·Software Engineer·Technical Phone Screen

Jul 2026

Anthropic software engineer interview with a graph traversal coding problem. Pretty standard on the surface but they pushed hard on the URL normalization angle which I wasn't fully ready for.

Write a function that crawls a website starting from a seed URL and returns all unique pages reachable within the same domain, using a provided helper that returns links for any given page.

“I went with BFS pretty quickly, which felt right.”

View Post

Anthropic·Software Engineer·Onsite - System Design / Architecture

Jul 2026

System design interview at Anthropic for a software engineer role. The question was a single massive open-ended prompt covering basically every dimension of backend architecture you can think of. Felt like they were stress-testing how far you could go rather than whether you'd hit a specific answer.

Design a scalable backend service that handles high-volume network I/O for millions of users, covering requirements, APIs, data models, capacity estimates, high-level architecture, scaling strategies, security, failure handling, observability, and edge cases like partial uploads or slow clients.

“This was the whole interview, basically.”

View Post