← Blog

Anthropic Coding Interview: One Team Just Dropped It

By the DevInterview.AI teamPublished November 3, 2026

On October 27, 2026, Anthropic engineering manager Felix Rieseberg wrote that he has removed the traditional coding interview from his team's hiring loops, arguing it no longer tells him anything useful now that Claude can solve small coding challenges faster than candidates. Anthropic engineering manager Felix Rieseberg says he has stopped using traditional coding interviews for his team, and as far as he's concerned, the coding interview is dead. The Register covered the post the next day, on October 28. One thing to be clear about up front: this is one manager's call for one team, not an Anthropic policy, and the company's general process still includes live coding.

What actually happened (and what did not)

Rieseberg's reasoning is narrow and worth quoting precisely. Small coding challenges in real code did tell us something meaningful about candidates. His point is that this signal has decayed, not that coding skill stopped mattering. He described the old format as a weak-but-useful predictor and said he does not expect it to return. "I don't think the coding interview is going to come back," said Rieseberg.

What did not happen: Anthropic did not abolish coding interviews company-wide. An Anthropic hiring manager drops the coding interview because it "no longer provides signal." Reporting is consistent that this is a single team's decision. Anthropic's careers page still lists live coding for technical roles and instructs candidates to complete those rounds without Claude unless told otherwise. If you are interviewing at Anthropic in the next few months, assume you will still write code in front of a human.

The move also is not happening in isolation. Larger companies are experimenting from the opposite direction: instead of removing the coding round, they are letting candidates bring an AI into it. The Next Web reports that Google is expanding coding interview pilots that let candidates use Gemini. That keeps the interview but changes the rules, so candidates work with an AI assistant during the round. At Google, the pilot has been framed around code comprehension rather than blank-page algorithm writing.

Why this matters if you're interviewing soon

Two trends are pulling in the same direction even though they look different. One team drops the coding round entirely; another keeps it but hands you an AI. Both are a bet that typing out a correct algorithm from memory is a weaker signal than it used to be, and that judgment, framing, and verification are stronger ones.

Here is our honest read, from running mock interviews every day: do not over-correct. Most loops you will sit in over the next six months still include a standard, human-graded coding round. Anthropic's general process does. Amazon, Google's non-pilot tracks, Meta, and the rest still screen on data structures and algorithms. The candidates who hurt themselves are the ones who read one headline and stop practicing coding. The right response is to keep your coding sharp and add the skills these new formats reward.

What to practice: coding, still

Treat the fundamentals as table stakes. The problems Anthropic has historically leaned on are not trick puzzles; they are messy, real-world-shaped questions that reward clean decomposition. From the DevInterview.AI question bank, the ones that come up most for Anthropic are good calibration:

Notice the pattern: none of these is an obscure dynamic-programming puzzle. They are problems where explaining your approach and reasoning about trade-offs is half the interview. That is exactly the muscle the new formats test harder. You can drill these and get feedback on practice.

What to practice: the signals that survive AI

If a coding round is being devalued or AI-assisted, the evaluator is spending more attention on everything around the code. Train these deliberately.

Problem framing. Before writing anything, restate the problem, list assumptions, name the inputs and constraints you are unsure about, and ask. In an AI-allowed round this is doubly important: a vague prompt to the tool produces vague code, and the interviewer sees both.

Trade-off reasoning out loud. Say why you chose a hash map over sorting, why BFS over DFS, what you would change at 1,000x the data. Narrate the decision, not just the keystrokes. We wrote a full guide on doing this without rambling; the short version is to state the decision, give one reason, and move on.

Verification. This is the skill the Gemini-style formats are built to measure. When an AI (or you) produces code, can you spot the bug, write the test that catches it, and reason about the edge case it missed? Practice reading code you did not write and finding what is wrong with it. Debugging under pressure is now a first-class interview skill, not a nice-to-have.

Evidence you can ship. Rieseberg's stated alternative leans toward real work. Have two or three projects you can walk through in depth: what you built, what broke, what you decided and why. Be ready to open the repo and explain a specific function. Our guide on talking about your projects covers how to structure that without turning it into a tour.

How to tell you're ready

You are ready for a standard coding round when you can solve a mid-level problem like Find Duplicate File in System while narrating your choices, and still finish with time to test. You are ready for an AI-assisted round when you can take a chunk of generated code, correctly explain what it does, identify at least one failure mode, and fix it, all while saying what you are checking and why. If you can only do the first, you are prepared for today's loops but not the pilots. Build both.

A simple weekly split that works: three sessions on timed coding (array, string, graph, and stack problems rotate well), one session reading and debugging unfamiliar code, and one session doing a full mock where you must explain every decision aloud. Company-specific prep helps too; our Google breakdown tracks what that loop actually asks.

FAQ

Did Anthropic remove the coding interview?

No. One engineering manager, Felix Rieseberg, removed it from his own team's loops as of late October 2026. An Anthropic hiring manager dropped the coding interview because it "no longer provides signal." Anthropic's company-wide careers page still lists live coding for technical roles.

Can I use Claude in an Anthropic coding interview?

Not unless you are explicitly told to. Anthropic's general guidance instructs candidates to complete coding rounds without Claude unless an interviewer says otherwise. Rieseberg's team change does not alter that company-wide instruction.

Is this happening at other companies?

Partly, and in a different form. The Next Web reports that Google is expanding coding interview pilots that let candidates use Gemini, which keeps the interview but changes the rules. Most large employers are moving more cautiously than Anthropic's single-team example.

Should I stop grinding coding problems?

No. The majority of loops you will face in the next few months still include a human-graded coding round. Keep your fundamentals sharp and add the newer skills: framing, trade-off reasoning, and verifying code you did not write yourself.

Sources

The real one is coming. Be ready for it.

Take a realistic AI-led mock interview with questions top companies actually ask, with live voice and real feedback.

Start a mock interview

Your first interview is free · no credit card required

Keep reading