Anthropic Manager Drops Coding Interviews: What Rieseberg Actually Said

A hiring manager at the company that builds Claude says Claude has made his coding interview pointless. Felix Rieseberg wrote on October 27, 2026 that "this month, I laid the traditional coding interview to rest," and that "as a hiring manager at Anthropic, I stopped asking candidates to do them." His post, titled "Hunger beats tenure," is part hiring note, part advice for junior engineers. It spread quickly because it comes from inside an AI lab. It is also narrower than some headlines make it sound.

This is a team-level decision by one manager, not an announced Anthropic policy. Below we cover what he wrote, what Anthropic's public candidate guidance still says, what the claim means for people applying to engineering jobs, and what you can do about it. For the company's own data on junior roles, see our post on how employment is changing at Anthropic.

TL;DR: what changed and what did not

What Rieseberg wrote

The central passage is short. He says small coding challenges in real code once "did tell us something meaningful about candidates," being a "weak but useful predictor" of success on his teams. Then: "That's no longer true. There's no more point in giving you a little React challenge than there is in quizzing people about assembly; Claude is now better at both. Our most impactful builders are successful regardless of how good they are at these tests."

He does not publish numbers. There is no data on hire quality, no before-and-after comparison and no sample size. It is a manager's judgment, stated plainly. That does not make it wrong, but it means the claim is an observation from experience rather than a measured result, and readers should weigh it that way. He also expects that "established engineers will still have strong reactions to its disappearance."

His X post, which links the essay, summarizes the same point: the coding interview "no longer provides signal," and he sees "an opportunity for junior engineers."

Hunger beats tenure: the argument

The post's optimistic claim is that AI coding tools reduce how much accumulated knowledge of frameworks and libraries limits impact. Rieseberg writes that he is "seeing engineers who are comparatively new fly past people who are probably better coders and have been around much longer." His reasoning is organizational, not technical. Companies, he says, are built so that a few people approve or block ideas, and in a world where execution is expensive you had to earn status before the organization would bet on you. "Execution isn't expensive anymore."

He supports the point with stories from earlier jobs, including Stripe, where founder Patrick Collison asked newcomers to pay attention to things that seem odd, and Notion, where a new hire asked for large dashboards that an older employee had assumed were impossible. The lesson is about fresh eyes: people who do not yet know why something is "impossible" often try it.

Cards pinned to a job board, one highlighted, representing hiring for engineers who build with AI

His five-step advice

1. Pick business impact over backlog. He cites Patrick McKenzie's 2011 line that engineers are hired to create business value, not to program things, and advises attaching yourself to the parts of a company that make money.
2. Pick a radical idea. Find a few ideas that would dramatically improve the odds of reaching the organization's goals.
3. Build it over a weekend. Small groups and solo engineers, he says, have flown past PRDs, buy-in meetings and lobbying designers by building a working prototype.
4. Focus on one good idea. Junior people often cram several ideas into one prototype and hide three good ones among twenty bad ones.
5. Lean into your fresh perspective. Newcomers empathize with users and have not learned the reasons things are "impossible."

His own example is internal. He says two relatively new Anthropic engineers, Anthony Morris and Alice Zhao, built a working prototype of a graphical interface for Claude Code over one weekend, and that the first public version shipped roughly eight days later. We cannot independently confirm those dates, but the product exists: our guides to Claude Code desktop parallel sessions and worktrees and Claude Code desktop mods show how it looks today.

What Anthropic's own candidate guidance still says

It would be easy to read the post as "Anthropic lets candidates use Claude to pass interviews." That is not what it says. Anthropic's public guidance on candidates' AI use, last updated July 10, 2025, tells applicants to complete take-home assessments without Claude "unless we indicate otherwise," and says live interviews are "all you" with no AI assistance unless stated. It encourages Claude for polishing a resume, researching the company and practicing answers, and forbids using it to invent experience.

The page also says Anthropic uses Claude for job descriptions, interview questions, candidate communications, hiring metrics, transcription and sourcing, and that it does not use candidate data to train Claude or let Claude make hiring decisions. Those lines are from mid-2025, and the page says the company plans to review them as capabilities change. Nothing we found shows an update tied to Rieseberg's post.

So there are two separate ideas. One is how a candidate may behave in a test. The other is whether the test is worth giving at all. Rieseberg is talking about the second.

What this means for people applying to engineering jobs

The honest reading is mixed. A manager at an AI lab says algorithm-style screens add little. But his replacement, scanning for "little prototypes that exhibit taste, empathy, and principled thinking," is harder to prepare for and rewards people who already have a body of work. It also favors candidates who can show finished things, which can disadvantage people without free time or access to tools.

Practical steps:

• Ship small things. One finished prototype that solves a real problem says more than a list of languages. Keep the scope to one idea, as he advises.
• Write down the why. A short README stating who the user is, what you chose not to build and what you learned is the "taste and empathy" evidence he says he looks for.
• Show how you work with AI. Use the tools, and be able to explain what you accepted, what you rewrote and how you checked it. Our Forward Deployed Engineer preparation guide covers a related role where this matters.
• Do not assume every company follows. Many employers still run live coding rounds, and some now test them with AI allowed. Read each employer's instructions, as Anthropic's own page recommends.
• Know the market. Entry-level pressure is real in some regions; see India's entry-level IT hiring and AI for a claimed-versus-verified look.

Two chairs and a clipboard with a check mark, an interview setting without a coding test

What this means for people who hire

If you run a hiring loop, the useful question is not whether to copy Anthropic but what your current test measures. A small coding challenge that a model can finish in seconds still tells you whether a candidate can read and reason about code, but it no longer separates strong from weak builders as cleanly. Options a team could consider include:

1. A review of past work, with a conversation about decisions and trade-offs.
2. A paid or time-boxed project where tools are allowed and the candidate explains the result.
3. A code review exercise where the candidate must find problems in AI-generated changes.

These are our suggestions, not Rieseberg's method; he says explicitly that he is still working it out and invites better ideas. Whatever you choose, document it, apply it consistently and check for bias, because portfolio-based hiring is easier to apply unevenly than a standard test.

Caveats worth keeping in mind

• One manager, one team. Rieseberg writes about "my teams." The source does not support reading it as the end of coding interviews at Anthropic or across the industry.
• No outcome data. The post offers no evidence that people hired without the test perform better or worse.
• The incentive. A lab that sells coding tools has reason to say coding skill matters less. That does not make the observation false, but it is context.
• "Junior" is a loose word. He says it is really for engineers "without social capital, status, or power."
• Survivorship in the examples. The prototype story is a success. Failed weekend prototypes do not get blog posts.

For broader context on how Anthropic employees and the public see the technology, see our look at the AI expectations gap between staff and public, and for the tool at the center of this story, our essay on why the terminal era is giving way to agents as the new primitive.

What to watch next

Watch whether other hiring managers at AI labs say the same, whether Anthropic changes its public candidate guidance, and whether anyone publishes data on hires made without coding screens. Those would turn an interesting opinion into evidence. Until then, the safest summary is that one Anthropic manager has stopped testing for what Claude does well, and is betting on what a candidate can build and why.

Related reading

• How Employment Is Changing at Anthropic
• Forward Deployed Engineer preparation guide
• Forward Deployed Engineer: the hottest tech role
• Claude Code desktop parallel sessions and worktrees
• India entry-level IT hiring and AI
• The AI expectations gap: Anthropic staff vs public

Quotes are from Felix Rieseberg's October 27, 2026 post and Anthropic's public candidate guidance as of October 28, 2026.