Comment by vunderba | Hacker News Reader

Comment by vunderba | Hacker News Reader

TLDR: Another article in favor of code reviews vs leetcode challenges.

About a decade ago during an interview, I was shown a piece of code (directly from their code base) which was about 50 lines long and we had a back and forth technical conversation in which the interviewer asked me the following questions:

- What does this code appear to do (both technically and more importantly in the context of the business domain of the company)?

- Are there are any red flags, edge cases, etc. that this code might not cover?

- If I were to rewrite it in the using modern libraries/tooling/best practices, what might I have done differently?

This is the same interview process that I now use on prospective applicants. In a single 30-60 minute dialogue it helps me answer questions such as:

- Can they size up and summarize existing unfamiliar code bases? In enterprise the ability to deal with legacy code is often more common than writing new code.

- Can they re-interpret old code using a modern lens? For example if we're discussing JS/TS (requires vs imports, callback hell vs async await, XMLHttpRequest vs fetch). For python (nested lists vs dataframes, simple numpy regression vs pytorch, etc)

- Are they able to communicate and convey their thoughts in a coherent and logical manner?

Yup, I went through a similar experience with a YC startup no less. First round was a code review with the CTO/co-founder and it was a web API and SQL schema with lots of little bits and varying "areas of interest" from data validation to exception handling (or lack thereof) to indices to type selection and other bits and pieces.

Left a very strong and favorable impression on me for how much more enjoyable it was as a candidate and also how much more it felt like he presented a sandbox within which we could discover and discuss.

Meanwhile there's a handful of companies (the one I work at included) using AI to automate code reviews.

Sure, these tools are not as good as a real human. Yet. Just like 3 years ago companies were starting to ban candidates to use GitHub Copilot, pretty soon candidates WILL find a way to either have ChatGPT open on the side, or listening to the operating system audio and providing real time suggestions during the interview (yeah I know this can be done today).

We either go back doing things on site, or we go crazy and let candidates use whatever they want during an interview. Want to use Cursor and Perplexity + R1? Sure, why not. This is what people are using at work anyway. This probably means interview conclusions are going to be even more nuanced, it can't simply be they solved the problem or not (and the interviewer liked the person), but during the actual conversation things moved organically to whatever direction the topic led to - which is what actually happens at work when pairing.

As you can see I have no idea how to "solve" this problem moving on, but one thing is for sure, the bigger the gap between tooling involved in actual work vs the interview, the bigger this problem is. It's one thing to ask people to write code on a whiteboard for 30 minutes (because they "write" code in a text editor), and another thing to ask people to write code in a whiteboard and the actual work is people chatting with AI Agents to produce so much software in 30 minutes that the entire thing would look like side projects 5 years ago.

Per a separate comment I made, I think you could add to this and call it an AI assisted code review. Not just can the user review the code shown, but can they review the code that AI creates?

Yes, AI code is sort of like any other code in that it can have errors, but I think taking it to the next step is whether they can actually modify or follow up with the prompt to get it to do the right thing.

AI is kind of like having a really smart and confident intern who's sometimes wrong or doesn't quite understand your request. But you're not a manager or a lead. You're just some junior hire who's got AI saying, "Hey I have this figured out and am very confident in my results" and it's your job to check their work.

I like this.

Reply on news.ycombinator.com