Jane Street adds design and LLM-focused rounds to its software engineering interviews

What Jane Street announced

Jane Street is reworking its software engineering interviews after years of leaving them essentially untouched. In a blog post that reached the front page of Hacker News, the firm said it began trialing new kinds of onsite rounds this summer for campus candidates. The new formats involve less hands-on coding and put more weight on software design, on judging whether code is well-structured and does what it was meant to do, and on using LLMs to produce code.

Most rounds, Jane Street stressed, will remain classic live coding sessions conducted without any AI assistance. Not every onsite includes the new formats yet, and candidates who encounter them will be told in advance what to expect. The rest of the hiring process is unchanged.

The firm kept the specifics deliberately thin. Because the formats are still evolving, it wrote, it would rather not publish a detailed description today of something that may look different six months from now.

Why the firm says the job has changed

Jane Street's own engineering practice drove the decision. According to the post, most of its engineers now use LLMs every day — not only to write code, but to read, debug and review it, along with tasks that have nothing to do with code at all.

Its interviews, by contrast, had stayed in the same shape for years: a candidate sits down with one or two interviewers, works through a programming problem, and narrates as they go. Jane Street says that format still tested the right fundamentals — problem solving, structuring work, communicating and collaborating — but that the balance of the job has shifted. With producing code now cheaper than it has arguably ever been, the firm argues other skills matter more: judgment about what is worth building, picking abstractions that hold up over time, and confirming that what was built is actually correct.

The post ties into earlier Jane Street writing on the same theme, including a podcast about how the firm uses AI in development and a piece on formal methods that touches on how agentic coding is reshaping daily work. It is signed by Matthew, an engineer who joined the firm in 2013 and works on distributed systems infrastructure and productivity tooling.

What candidates do and do not need

Two reassurances stand out for applicants. First, deep experience with LLMs is not a requirement. Candidates should not feel they need a polished methodology for working with AI agents to do well, the firm says, because what it is ultimately looking for is people with strong instincts for building software — and it does not believe it needs an LLM to evaluate that.

Second, there is no expected setup. Engineers have folded these tools into their workflows in different ways and to different degrees, the post notes, and nobody is expected to use them for everything. What matters is thinking clearly about the tools themselves: what they do well, where they fall short, and when to reach for something else.

Interviews for more experienced candidates will continue to be more varied, mixing hands-on software production, system design and discussion of prior work, as has been the case for some time.

Why it matters

Jane Street is a closely watched engineering employer, and changes to its hiring process read as an industry signal. The shift is a concrete data point in a wider argument: as code generation gets cheap, the scarce skills move toward design judgment, verification, and knowing when to trust machine-written output.

At the same time, the firm's restraint is notable. It is keeping most rounds LLM-free and explicitly not requiring AI fluency, suggesting it still sees unaided problem solving in front of an interviewer as its most reliable signal. For candidates, the practical takeaway is straightforward: keep preparing for traditional live coding, but expect interviews to increasingly probe how you evaluate code — and the tools that write it.