Comment by johnfn
7 hours ago
It's funny because this repo really does seem vibe-coded. Obviously I have no reason not to believe you, but man! All those emojis in the install shell script - I've never seen anyone other than an AI do that :) Maybe you're the coder that the AI companies trained their AI on.
Sorry about the job interview. That sucks.
There's even a rocket emoji in server console.logs... There are memes with ChatGPT and rocket emojis as a sign of AI use. The whole repo looks super vibe-coded, emojis, abundance of redundant comments, all in perfect English and grammar, and the readme also has that "chatty" feel to it.
I'm not saying that using AI for take-home assignments is bad/unethical overall, but you need to be honest about it. If he was lying to them about not using any AI assistance to write all those emojis and folder structure map in the repo, then the CTO had a good nose and rightfully caught him.
As a big believer in documentation and communication in general, there's this inevitable double-bind that people hate whatever you give them and also hate it if you give them nothing. LLMs have made this worse.
No emojis and any effort to be comprehensive? Everyone complains "what is this wall of text", or "this is industry not grad school so cut it out with the fancy stuff" or "no one spends that much time on anything and it must be AI generated". (Frequently just a way of saying that they hate to read, and naively believe that even irreducibly complex stuff is actually simple).
Stuff that's got emojis, a friendly casual tone and isn't information dense? Well that's very chatty and cute, it also has to be AI and can't be valuable.
Since you can't win with docs, the best approach is to produce high quality diagrams that are simultaneously useful for a wide audience from novice to expert. The only problem is that even producing high quality diagrams at a ratio of 1 diagram per 1k lines of code is still very time consuming to produce if you're putting lots of thought into it, double so if you're fighting the diagramming tools, or if you want something that's easy for multiple stakeholders with potentially very different job descriptions to take in. Everyone will call it inadequate, ask why it took so long, and ask for the missing docs that they will hate anyway!
On the bright side, LLMs are pretty great at generating mermaid, either from code, or natural language descriptions of data-flows. Diagrams-as-code without needing a whole application UI or one of a limited number of your orgs lucid-chart licenses is making "Don't like it? Submit a PR" a pretty small ask. Skin in the game helps to curbs endless bike-shedding criticism
> No emojis and any effort to be comprehensive? Everyone complains "what is this wall of text", or "this is industry not grad school so cut it out with the fancy stuff" or "no one spends that much time on anything and it must be AI generated". (Frequently just a way of saying that they hate to read, and naively believe that even irreducibly complex stuff is actually simple).
> Stuff that's got emojis, a friendly casual tone and isn't information dense? Well that's very chatty and cute, it also has to be AI and can't be valuable.
As a counterpoint, I can confidently say that I've never once had anyone give any feedback to me on the presence or absence of emojis in code I've written, whether for an interview, work, or personal projects, and I've never had anyone accuse my documentation of being AI generated or gotten feedback in an interview that my code didn't have enough documentation. There's a pretty wide spectrum between "indistinguishable from what I get when I give an LLM the same assignment as my interviewee" and "lacking any sort of human-readable documentation whatsoever".
Oh my god Becky, there's even a rocket emoji in the server console logs!
Should I also be "honest" about tab-completion? Where do you draw the line? Maybe I should be punished for having an internet connection too. Using AI for docker/readme's/simple scaffolding I would have done anyways? Oh the horror!
There was no lying because there was no discussion or mention of AI at all. Had they asked me, I'd have happily told them yes I obviously use AI to help me save time on grunt-work, I've been doing this stuff for like 15 years.
It's an unpaid take-home assignment. You'd have to be smoking crack to think that I would be rawdogging this. Imagine if I had a family or a wife or an existing job? I'd dump them after getting linked their assignment document.
Honestly at this point in the AI winter if you are a guy who has AI-inspired paranoia then I don't want to work for you because you are not "in the know".
> Should I also be "honest" about tab-completion? Where do you draw the line?
I'd probably draw it somewhere in the miles-long gap between tab completion and generating code with an LLM. It sounds like that's where the company drew it too.
You have that you’re the founder of an AI company in your hacker news profile, and your take home looks completely vibe coded. Why in the world are you surprised that a hiring manager is a little suspicious about your coding skills?
Given what you’ve said in your other comments, it seems like you used AI in a way that I wouldn’t have a problem with but just briefly looking through I can see how it would look suspicious.
1 reply →
I used AI for the Docker setup which I've already done before. I'm not wasting time on that. Yeah you can vibe code basic backend and frontend and whatnot, but you're not going to vibe code your way to a full inverse kinematics solution.
I'm not a math/university educated guy so this was truly "from the ground up" for me despite the math being simple. I was quite proud of that.
So what was the issue the CTO had with vibe coding? Had you disclosed to then that you used LLMs for coding "basic" features outside the math and whatnot?
CTO's previous job was at Palantir, perhaps he has some reasons to be paranoid
The hiring manager told me that they were getting a lot of "signal to noise" ratio in terms of their hiring, where they'd bring someone on-site who had a good assignment and apparently more often than not, these candidates would shit the bed in a live environment. So the CTO made a live take-home assignment and didn't tell anyone. I was told that he did this to weed out the low signal-to-noise people they dealt with recently.
>Had you disclosed to then that you used LLMs for coding "basic" features outside the math and whatnot?
No it seems completely immaterial. I'll happily talk about it if asked but it's just another tool in the shed. Great for scaffolding but makes me want to rip my hair out more often than not. If it doesn't one-shot something simple for me it has no use because it's infuriating to use. I didn't get into programming because I liked writing English.