Hacker News Reader

Comment by slt2021 | Hacker News Reader

slt2021

Kudos to you for doing this.

However, I will retire from this cursed industry if this will be the expectation in the future

Can't wait for 2035 when we’re debugging the prompt queue pulling data from the prompt lake, while and resolving issues in the contex window eviction service, all while the team is 90% percent vibe coders with no coding knowledge introducing more bugs than features.

Tbf, some colleagues I've had were introducing more bugs than features just fine before LLM's were even a thing.

I've once been at company that had 90%+ such colleagues.

Uff, if that is the future of this industry, I'll retire as well

At least you can correct them, right? Imagine working with pure vibe coders with no CS degree or even a bootcamp under their belt.

I've had one co-worker with something like a decade of experience on paper, who was proud of his C++ despite having never heard of the standard template library — lots of `new` and `free`, not a single smart pointer (https://en.cppreference.com/w/cpp/memory#Smart_pointers). And the code they wrote had a lot of copy-paste going on, which I ended up finding because I'd put in a "TODO: deduplicate this" comment somewhere and found it in his newly duplicated class one day.

They absolutely were not interested in learning anything. I left knowing more C++ than they did despite having started there with total C++ experience of a hello world tutorial, and the fact that I still don't count myself as a C++ dev today.

To be fair when a company says they use C++, it can mean anything from "C with classes" to crazy metaprogramming with almost automatic memory management. Since they have over 10 years experience, they are almost definitely in the former camp.

I would never utter the phrase "I know C++" because it can mean so many different things to so many different people, and I don't think anyone truly knows the whole language.

Not using templates nor smart pointers doesn't sound that bad to me(unless the entirety/majority of the codebase was written with them in mind), the duplication thing is more questionable.

It's not so much that this specific person didn't use smart pointers, it's that they had never even heard of them, and wasn't interested either.

"C with classes" is probably a good description, given what I saw from that one person — they didn't understand sub-typing either, and only had a cargo-cult understanding of access specifiers (revealed when the rest of us asked them why they'd duplicated a class file rather than subtyping).

Tbh, I also (sort of) knew C++, studied in school and a few semesters worth in college (CUDA, DSA, Computer vision elective,compiler design) but I still don't know STL. (I had been then interviewing using Java and Python.)

Nope! I reverted a commit once, since a colleague pushed something that didn't compile to master. Sent the guy a polite message notifying him that something seems to have been amiss with his last commit, and to please let me know if he wanted help fixing it.

Boss called me 5 minutes later and tells me off for creating "bad vibes" in the work environment.

Colleague then proceeded to forcepush his "fix" that still didn't even compile to master, removing a new feature I was about to roll out to production, because he didn't know how to merge his changes with the revert commit I'd added

This was when I decided to quit

Oh I should add this developer bragged he had 10+years working experience. Not that I believe him, but still

I know some talented coders who were doing quite well before. Now they fallen into vibe coding and when I come across a bug they just introduced and I can’t seem to find the source they reply they have no idea but will have a look.

The decline in the skills are clearly visible. And they’ve only vibe coded under a year.

Can say so for myself. Have been hitting LC lately for an upcoming interview and I have found I have gotten worse like considerably worse, after having grinded in college and barely touching it for 6 years. I had to look up how to implement topological sorting today for example and even then flubbed it.

Because it's somthing you never need to implement in any real-world job, unless that job is developing a library routine to do topological sorting.

I dont agree. The reason you are forced to learn DSA in college and is tested in LC is because these data structures and algorithms are everywhere.

You may claim that nobody ever will need to know about topo sort, but keep using AirFlow for your pipelines or storing and display a Sitemap tree on your website.

if you dont know the basics, you will inevitably reinvent in using substandard, inefficient data structures and buggy algorithms.

I have not thought about writing a sort or any kind of complex data structure beyond a dict or an array/list since my undergrad CS days which was almost 40 years ago. It just doesn't come up. If it does in your job, sure you need to know it. For most jobs it doesn't.

Agree. Topological sorting isn't a part of my job. A hash map might be the most complicated data structure I've used on the job. (Relatively speaking)

Me and another tech lead recently got into an argument with one of our "senior" devs about how to best implement logging in a new service. They started sending screenshots of text to bolster their points. When asked where that was coming from they admitted they were just asking chatgpt. It was infuriating.

I had once an incompetent manager provide a feedback on my code using ChatGPT.

This is the BIGGEST red flag of a person "faking till they make it"

I guess we could write a whole thread on incompetent managers, I used to work with a manager who's only talent is rephrasing something someone else said in a more serious way, or just telling something in a whole list of buzzwords that make him look he's talking something serious. In-fact when it can be summarized in few simple few words. Apart from that he had zero technical knowledge, and I still wonder how he came to that position

I would like to think that no company that actually wants to make money selling products to users, rather than stories to VCs, will do this.

Its a joke...but I do know "vibe coders". There are some recent grads I know who have supposedly studied programming yet I don't think they write a single word of code themselves and get confused with simple concepts like reading a simple database even with select * statements.

How is this any worse than the current system where your resume is just keyword-filtered? It seems like a straight upgrade for my resume to be discussed by agents that know the difference between Java and JavaScript and aren't going to pass on me because my resume didn't explicitly mention 'scrum' and 'agile' as skills.

For me, the whole point of the resume is that the applicant has highlighted those parts of their experience they consider relevant. I then pick the points that look most interesting to me and go really deep in them in the interview. So, ok, leave that to an LLM I guess? But don’t be surprised if I go super deep on something you weren’t expecting.

Edit: like for instance, if you slip up and put C++ on your resume, I will drill you on it unmercifully. In my experience 19 out of 20 people who list C++ experience can’t compile their way out of a wet paper bag.

I think I feel ya on some level but I also think that when the process is refined it will be much less exhausting to update our resumes with the help of an LLM. Underneath this tool is just consuming the data I already present to the world through my website, resume, linkedin, and github.

Once there is a "one click connect to an MCP server" workflow this type of thing will make more sense for this type of use case, but right now how would you say this improves on the status quo of a resume PDF you can upload to your AI chatbot and ask questions about? Aside from demonstrating your own proficiency with MCP tech, that is. I ask because the current amount of work and tech knowledge required is greater than it would be for the PDF-based workflow, but I might be missing something.

Edit: there was an example in another answer, "I offer a contact tool when you use the hosted server because I can hide away my email credentials and expose a way for the LLM to send me an email."

The standard PDF resume is optimized for the human to read. The information density there is pretty low. Take a look at https://ai.jakegaylor.com/llms.txt and compare that to https://jakegaylor.com/JakeGaylor_resume.pdf

Now we can spend our time more on the content and less on the presentation.

You can already use claude desktop, upload your resume, point it to your website, paste in some stuff from linkedin and output an llms.txt. You can get 80% of the way with just a couple of clicks.

It's the modern version of "Have your answering machine call my answering machine!"

> I offer a contact tool when you use the hosted server because I can hide away my email credentials and expose a way for the LLM to send me an email.

Yeah, but this is the modern equivalent of the "Stavros at Gmail dot com", it's basically antispam by obscurity. Just wait for one spammer to send three seconds writing something that will parse emails from all your MCP commands and that's defeated.

How is this not better for engineers than having to maintain a LinkedIn page or a PDF-based resume?

Maintaining a PDF-resume takes minutes.

My resume is just a chunk of HTML with `size: A4`, takes literally seconds to update it as it's just simple HTML and the "export" process is just ctrl+p in any browser and saving as PDF.

Reply on news.ycombinator.com