I Messed Up My Google PM Vibe Coding Interview | Hacker News Reader

I Messed Up My Google PM Vibe Coding Interview

old.reddit.com

61 pointstaubek45 comments

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.

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.

Seems a bit off. Big tech interviews are very structured, the recruiter should tell you exactly what will be assessed. Might be a fake post.

recruiters in big tech can't screw up or be incompetent?

I look forward to the day these vibe coded vapid applications are in prod and everything collapses

Seems like anyone surprised by this isn’t tracking what big/biggish tech product leaders are already moving toward; the roles of PM, UX/design, and POC dev are going to blur dramatically over the next 1-2 years.

Eventually, a solid CSM should be able to hear a customer pain point, spit out a quick POC, and begin a segmented rollout to validate it as quickly as possible.

Also agree with suspicion around this post though… big tech interviews, especially in AI adjacent roles, are going to be structured well enough that this probably wasn’t actually a surprise. I’d much more believe the reverse: “I was told I would have a technical/vibe coding component and didn’t!” (much less exciting headline of course)

Absolutely! In fact, this used to be the norm before fhe MBBs and ex-IBs who did an MBA flooded the PM market in the 2012-15 period thanks to Google's PM org.

PMs who started before 2012 or after 2022 tend to have a product mindset ("I'm building a product to solve a problem").

This requires domain experience - you can't learn this from an MBA. You need to have started of working in that field in order to become a truly strong product manager. Before the 2010s, most MBA PMs tended to be staff or principal engineers, sales engineers, or support engineers sponsored by their employer to attend part-time MBA programs like Berkeley Haas [0] or Stanford HCP MS&E [1]as a finishing school and return as a business minded engineer.

Google and Twitter (back under Dorsey in 2010-11) changed the whole PM hiring process industry wide by prioritizing MBB personas and MBB-style interviews as one of their heads of PM at the time was a former Partner at McKinsey and brought the McK process into the tech industry, despite more product minded people like Salar, Marissa, and Sundar helping build core fundamentals of what became the Google behemoth (Xooglers, please correct me if the history is wrong - it's been 15 years and I do think I messed up some of the chronology).

[0] - https://ewmba.haas.berkeley.edu/

[1] - https://msande.stanford.edu/academics-admissions/graduate/ms...

how do you think role of an engineering manager will change?

Hmm... reading between the lines, here, it looks like Google is out to save a bit of dosh. Instead of hiring a PM and some developers, they just hire a vibe-friendly PM.

Sign of the times, I guess...

Based on the current state of AI that's strange. They must be 100% sure that vibe coding will eventually replace engineers for them to go in that direction. Which is also strange because i don't see it still. Isn't it too early to replace engineers with vibe coders

not if you're trying to impress your bosses to get a promotion.

Exactly, you align your business objectives with the vision of your managers and the industry trend.

Right now it appears that most industry leaders have fully committed to AI. It would be foolish (career wise) to speak openly otherwise.

what's next. Companies will just let the customers vibe code what they're asking for

If there’s something worse than an average engineer doing vibe coding, that’s a pm doing vibe coding. What’s next? Marketing guys automating backups?

Thanks for the chuckle - you know there's more than one marketing guy out there who will say in all seriousness "I can do that for you" :-)

Or just a technical product manager.

Honestly, the biggest mistake IMHO was hiring non technical product managers and putting them in charge of technical teams. They’re almost always annoying af, useless, and tbh “people skills” only go so far because it’s mostly a systems problem that communication falls apart.

Having worked at successful tech companies (that IPO’d), I have never met a non-technical product manager who was any good or worth remembering.

The best thing happening in tech right now IMO are the layoffs affecting PMs and other "manager" types. We all know they add very low value because their actual role is "scapegoat" when their projects fail (mostly due to them in the first place). It is also why they were almost always the only ones fired prior to AI revolution.

Vibe coding itself is such a highly variable thing, it’s stupid to have been expected to do anything in particularly brilliant in such a situation.

The interviewer is an idiot.

The whole point of vibe coding is you are asking for what you want and the AI gives it to you. If you don’t get what you want, there are a bunch of reasons why that could be, but many of them are only available in hindsight.

I'd be genuinely curious to see a tool chain that actually makes 'vide coding' work.

I currently run a set of docker containers for proxy, backed, frontend, and authserv, and the best I've got is "edit" mode in vscode...and while it probably 2-3x my output, it maybe 0.8x my quality, and 0.5x my understanding/scalability.

Looks like they want to hire a PM to build a vibe coding software. And they want a PM that understands the usecase.

If the post is to be taken at face value, they weren't "vibe coding", they were trying to implement an agent:

> Instead, I jumped into building my Langchain backend agent first and completely messed up the code for tool calling in Langchain.

Vibe coding involves the agent existing before you start.

The author thinks they messed but there’s no evidence of that. If the goal was to demonstrate building something with ai the interviewer shoul be able to see past the stress to the skills demonstrated..if they are willing.

I distinctly remember having to agree to some non-disclosure when I ventured this way a long while ago

From what a couple of other commenters mentioned, the post might just be a red herring, so there's nothing actually being disclosed...

Don't worry about it. I recommend not working for Google, Apple, Microsoft, Amazon, or any major tech company. You are disposable at these companies. Rather than take one for the team and put back some of their ridiculous salaries, they'll throw you off the boat and gaslight you to believe you are just inadequate.

Nay. I found education, healthcare, and other industries that are not tech focused to be much more rewarding. You are the expert in those industries and are given a chance to lead.

I build ML for healthcare orgs. Different things. An immunology prize needed a model to filter out applicants because they were overloaded. A nurse staffing agency wanted to measure burnout in nurses using the data they had. A speech therapy provider. A school district. And on and on.

I'd rather work for them than bend the knee to Tesla's Technoking.

Actually, there's a rumor that if you punch the world's richest man in the face, his wealth rubs off on you. You will spend some time in jail, but you will be rewarded. Make sure they catch it on camera.

What's the ranting about Musk got to do with the first part of your post?

I expect this to become standard for many roles. This is the future. It's hard to say when or how but for anyone who has spent > 6 months, interacting with LLMs on a daily basis, it's clear this is the way.

Extremely fake like every Reddit post nowadays.

Reply on news.ycombinator.com