Report · felixrieseberg.com
Hunger beats tenure: Advice for junior engineers
felixrieseberg.com
Source published October 27, 2026
Summary: 1 min read
53/100
Summary
Felix Rieseberg, a hiring manager at Anthropic, says he stopped using traditional coding interviews this month because Claude now outperforms candidates on small React challenges and assembly questions. He says these tests once weakly predicted success but no longer identify the most impactful builders, and that engineers who are newer to the field can surpass more experienced programmers when they use AI effectively. Rieseberg argues that junior engineers, or engineers without organizational status, should focus on business impact rather than backlog tasks. He recommends understanding an organization’s goals, proposing a radical idea tied to those goals, and building a focused working prototype over a weekend instead of waiting for formal requirements, approvals, or cross-functional support. He says prototypes should communicate one strong idea clearly and demonstrate that their creator can lead it. Rieseberg cites an Anthropic example in which relatively new engineers Anthony Morris and Alice Zhao built a graphical interface prototype for Claude Code over a weekend; the first public version shipped roughly eight days later. He says he is now looking at resumes and portfolios for small prototypes showing taste, empathy, and principled thinking, while still trying to determine how to interview people who build effectively with AI.
Key Takeaways
- Felix Rieseberg says he ended traditional coding interviews for candidates on his Anthropic teams because Claude is better at the coding tasks those interviews typically test.
- Rieseberg says AI lowers the cost of execution, allowing engineers with less tenure to demonstrate impact through working prototypes.
- He recommends that junior engineers connect ambitious prototype ideas to business goals and focus each prototype on one clear opportunity.
- Anthropic engineers Anthony Morris and Alice Zhao built a graphical interface prototype for Claude Code over one weekend, and its first public version shipped roughly eight days later.
What the discussion said
The thread largely ignored the article’s broader career sermon and zeroed in on how AI is scrambling junior-engineer hiring. Commenters agree that the old closed-book coding puzzle is losing its claim to measure real work: a candidate can now hand a problem to a chatbot, so interviews need to test how someone frames ambiguity, steers AI output, extends a system, and explains its weak spots. Practical alternatives drew interest, especially building a modest product in a familiar environment with AI available, then probing architectural choices, or reviewing realistic domain code. But enthusiasm for abandoning standardized interviews is tempered by distrust of subjective judgment. Several readers argue that Leetcode’s flaws are at least visible and consistent, whereas open-ended assessments let interviewer mood, cultural fit, and bias decide the result. They also reject weekend experimentation as a supposedly meritocratic signal when it rewards candidates with spare time and shifts company risk onto workers. AI’s productivity boost drew a similarly sour response: it may enable faster exploration and more parallel model experiments, but organizations can convert those gains into leaner staffing and heavier workloads rather than protected time for learning. The Anthropic perspective was treated as unusually privileged, not a template for ordinary firms.
Where opinion split
The sharpest dispute is whether AI should end standardized coding interviews. Supporters say chatbot-assisted, open-ended work better captures modern engineering because candidates must direct tools, reason about architecture, and navigate incomplete requirements. Skeptics say removing normalized tests replaces an imperfect but legible benchmark with interviewer discretion, creating more room for bias and arbitrary gatekeeping.
Coverage & discussions
Hacker News · felixrieseberg.com
Community discussions
Community Sentiment
Positives
- AI-assisted practical exercises can reveal whether candidates can turn model output into a coherent, extensible system instead of merely reciting puzzle solutions.
- Fast AI-enabled prototyping lets small teams test multiple language-model designs in parallel, killing weak approaches quickly and concentrating effort on the promising one.
- Open-ended interview problems may better reflect real AI-era engineering, where finding and validating information matters more than memorizing obscure algorithms.
Concerns
- Chatbots make unsupervised Leetcode answers a weak signal, forcing hiring teams to redesign evaluation before automated help turns every puzzle into a copy-paste exercise.
- Replacing standardized coding tests with open-ended judgment risks trading a flawed common yardstick for cultural-fit theater and interviewer bias.
- Weekend AI experimentation shifts failed-project risk onto junior workers and favors applicants with abundant free time rather than stronger engineering ability.
- AI productivity gains may become a rationale for layoffs and intensified workloads, not paid space for engineers to learn and explore.
Related Articles
The ladder is missing rungs – Engineering Progression When AI Ate the Middle
Mar 30, 2026
Will Claude Code ruin our team?
Mar 8, 2026
Vibe coding and agentic engineering are getting closer than I'd like
May 6, 2026
We might all be AI engineers now
Mar 6, 2026
AI doesn't generate working products, that's still your job
Aug 1, 2026