Our software engineering interviews are changing a little
A high-profile trading firm’s blog post on tweaking its software engineering interviews for the LLM era prompts debate over what modern technical hiring should measure. Commenters argue about allowing or even expecting AI use in coding rounds, how to test genuine understanding versus prompt-driving, and whether traditional leetcode-style questions or git “basics” still make sense. Others note the company’s frequent recent appearances on Hacker News, questioning whether this reflects organic interest, deliberate recruiting-oriented marketing, or simple statistical clustering of posts.
Perception of Jane Street’s Presence on HN
- Many notice an unusual spike in Jane Street posts, including multiple front-page stories in the same week.
- Some suspect deliberate marketing or astroturfing, especially to attract engineering talent.
- Others argue it’s normal “halo effect” and Baader–Meinhof: more visibility leads to more upvotes; data shows steadily rising mentions over several years.
- Jane Street’s high pay, “elite” reputation, OCaml use, and tech talks are seen as strong draws for HN readers.
Reactions to the Interview-Changes Post
- Several readers find the blog post vague and light on specifics; some see it as thinly veiled recruiting content.
- Others think flexibility around LLM use in interviews is reasonable given changing workflows and that interview formats must adapt.
How to Interview in the LLM Era
- Many argue traditional live-coding / leetcode is now even less meaningful, since LLMs can solve those problems easily.
- Proposed alternatives include:
- Letting candidates use LLMs but requiring deep explanation of generated code.
- Focusing on architecture/system design, debugging, code review, and ability to guide or correct an LLM.
- Intentionally using a weaker or mistake-prone LLM to test whether candidates detect and fix issues.
- Some foresee future interviews testing “AI orchestration” skills: setting goals, guardrails, verification, and workflows around agents.
Debates About Tools: Git, PRs, and LLM Reliance
- Long sub-thread on asking candidates to open a PR as a “git basics” test.
- Pushback that PRs are about hosting platforms, not git itself; many capable devs rarely use PRs or specific forges.
- Broader split between people who insist engineers should know core git operations vs. those who see LLM/IDE assistance as fine for infrequent tasks.
Broader Critiques of Interviews and Hiring
- Multiple comments question whether any of these interview rituals are evidence-based or predictive of job performance.
- Some advocate simpler conversational interviews or short trial periods instead of elaborate multi-round processes.