Hacker News Reader

Comment by abxyz | Hacker News Reader

Buried in the comments the OP acknowledges that this was specifically an AI role and that the OP was unprepared to be asked to vibe code in the interview, not that the OP was surprised the role itself involved vibe coding. We are reading it as if this was a PM interview and vibe coding was sprung on the OP — and that’s why it is interesting — but it was actually just someone who didn’t expect an interview for an AI role to involve demonstrating the skill.

View on HN

I still find it bewildering that vibe coding would be considered a core competency for a PM.

I have no idea if this is why they asked about it, but I can imagine it could be a great skill to create functional prototypes to share with actual devs.

or replace actual devs

Certainly, that is the the long term goal. I am just some schmoe, but I have been thinking for quite some time that non-top 10% devs should really start to think in product terms to be prepared for this future. I know this is heresy.

I find it bizarre that it's considered a competency for anyone.

The sad thing is everyone will follow this interview practice like they did before with other ridiculous questions

I find it bizarre that people can be bothered to write code by hand any more.

It's still quite necessary for everything beyond junior grade code.

Which, for PMs trying to test MVPs and write deep specifications, is exactly what you need.

I certainly can be bothered, because it’s great fun. But when Claude Code can do the same thing 20% faster, I’m a fool for not applying that.

The trick is: when.

Not always.

And hardly at all in “yolo mode”.

It's nice if you enjoy coding, I agree. I don't enjoy it, so I just have Claude do everything. I have to review all the code it writes, of course, or at the very least the function signatures and general architecture, but I can't be arsed looking up API/library docs myself ever again.

My current employer is actually getting ready to do required vibe coding training. Probably not a good time to be a tech company shareholder tbh.

depends on which tech company

If you are the backbone of any major LLM it is.

I wouldn't call it bewildering. We're already seeing some capacity of models to take PRDs and turn them into PoCs. We're not far from being able to literally test PRDs by how well the models implement them. In other words, how well can a PM write inputs makes a difference. And we're close to having quasi instant "scoring" for that.

The whole point of hiring a PM is to have an SME who truly UNDERSTANDS customer and user painpoints.

If you are the PM for building AI products, you better dang know how to use AI productivity products.

The era of MBB and process driven PMs is dead. We are back to the pre-2013 era of PMs as domain experts first and foremost.

As an ex-PM, thank goodness.

Process oriented PMs are useless. Process is a tool, not a product competency. UNDERSTANDING who a customer is and the problem they are facing is what matters.

We build products and companies in order solve a problem, not the other way around. PMs and Founders with this mindset tend to succeed from an investment standpoint.

It makes sense that they need to have good skills with AI but vibe coding also requires coding skills of a high enough ability to be able to quickly assess the quality of what you're seeing.

PMs are increasingly expected to have some technical sharpness (including understanding code) - not enough to be a technical IC, but enough that they can create their own lab environment and give architecture feedback.

This is especially prominent in much more technical products like AI/ML, Cybersecurity, Data Lakes, etc.

If you are building and selling a technical product to technical personas, you better understand the pain points which only comes from having actually experienced that.

The Emperor's New Code.

Unfortunately there is a type of product manager who vibe codes enough to think that most software engineering problems are trivial solvable with AI and trys to push product direction towards AI solutions. Like using overwrote AI-powered browser agents instead of simple screen scrapping. These types of PMs are good at talking to leadership but tend to talk to engineers like they are prompting a LLM.

Reply on news.ycombinator.com