Applicants are using AI too. I've heard from people who hire/post jobs that they gets hundreds to low thousands of applications, and maybe 5% of them have any relevant experience. The problem is the breakdown of trust is costing all of us.
> Applicants are using AI too. I've heard from people who hire/post jobs that they gets hundreds to low thousands of applications, and maybe 5% of them have any relevant experience.
This happened before "AI" too. When all it takes is clicking an "apply now" button on LinkedIn some desperate people will spam any job they see.
I recall seeing one where you had to send a specific payload to an https endpoint to apply (or it might have been an automated screen immediately after the application was submitted). Forcing potential candidates to briefly open the curl manpage seemed like a similarly elegant solution to me. I doubt it works as well in the era of LLMs though.
similarly, i remember at least one organization (pre-Songtradr Bandcamp, i think) who didn't publish some of its open technical roles anywhere except in HTML comments on their website. they only wanted to attract folks who liked to poke around and look under the hood.
Snail mail has started to break down in the USA. I remember when I was a child letters always took 3 days to be delivered. Now I've sent letters to family members that took more than a week to arrive. I imagine that makes it hard for a candidate to plan or align interviews.
As far as I can tell it costs almost 10 times as much to send a letter certified mail (or any other option with tracking). And it means I can't just use a regular stamp, I have to go to the post office or use a third party service like stamps.com and print out a label.
And in some places they are incentivised to do so, as they may need to prove a certain number of applications per-week, or they'll lose unemployment benefits, so they end up applying to all sorts of unsuitable stuff.
At this point, we think using AI and being able to use AI effectively is a skill in and of itself. When you're hired, you'll have access to AI. You'd be expected to be able to use said AI effectively.
So, we still give you a FizzBuzz. You can use AI. Even if we told you not to use AI, we know almost everyone would use AI. But you have to understand the FizzBuzz and be able to explain it to us and make changes to it "live". The amount of people that get weeded out just by having to explain the code they "coded themselves" is staggering (even pre-AI, even on a take home where you had no "OMG I suck at live coding" pressure).
It's been a year since I've actively given out take-homes for hiring, but I'm not sure I agree that everyone will use AI. I designed half the questions to be impossible for current-gen AI to answer without the candidate actually knowing what's going on [0], and only ~1% of candidates who responded did poorly on that half and not the other half (and, if we're worried about LLMs being better than I think, not all that many candidates passed most questions either).
[0] The most reliable strategy I've found for that is choosing questions where the wrong answer is the right answer for some much more common question. Actually spending a few seconds and solving the problem easily lets a human pass, but an LLM with insufficient weights or training data (all of them) doesn't stand a chance.
Thanks for clarifying - I kinda get the idea but would love to see an example for this.
I’ve mostly given up on all of the standard techniques for interviewing sadly, just because “using ai” makes a lot of them trivial, and have resorted to the good old fashioned interview, where I screen for drive, values and root cause seeking, and let people learn tech/frameworks/etc themselves.
But I was wondering, isn’t a take home question still good, if you give a more open ended and ambitious task, and let people vibe code the solution, review the result but ask for the prompt/session as well?
People will be doing that during normal work anyway, so why not test that directly?
One such question (obviously tailored to the role I'm hiring for) is asking whether SoA or AoS inputs will yield a faster dot-product implementation and whether the answer changes for small vs large inputs, also asking why that would be the case.
I typically offer a test with a small number of such questions since each one individually is noisy, but overall the take-home has good signal.
> why not test that directly?
The big thing is that you don't have enough time to probe everything about a candidate, especially if you're being respectful of their time and not burning too much of yours. Your goal is to maximize information gain with respect to the things you care about while minimizing any negative feelings the candidate has about your company.
I could be wrong, but vibe coding feels like another skill which is more efficient to probe indirectly. In your example, I would care about the prompt/session, mostly wouldn't care about the resulting code, and still don't think I would have enough information to judge whether they were any good. There are things I would want to test beyond the vibe coding itself.
In particular, one thing I think is important is being able to reason about code and deeply understand the tradeoffs being made. Even if vibe coding is your job and you're usually able to go straight from Claude to prod, it's detrimental (for the roles I'm looking at) to not be able to easily spot memory leaks, counter-productive OO abstractions, a lack of productive OO abstractions, a host of concurrency issues LLMs are kind of just bad at right now, and so on. My opinion is that the understanding needed to use LLMs effectively (for the code I work on) is much more expensive to develop than any prompt engineering, so I'd rather test those other things directly.
Yes, that's why I said, we have you explain what you "vibe coded" and then also do an actual live coding part where you have to make further changes. Via screen sharing.
The amount of people that can't even navigate "their own" code is astonishing. Never mind explaining what it does or making changes.
For the 95% irrelevant and 5% relevant groups, I wonder what percentage of resumes come in through a third party recruiter.
I get tons of spam that could be generated by even a basic LLM based on public information about me, but for positions that are not a reasonable fit.
Apparently, it is common for such cold calls to come from “recruiters” that are not affiliated with the hiring firm, but are trying to collect some sort of referral bounty.
I have no idea why an HR department would be dumb enough to set up such a pipeline (by actually paying for the third party “service”), but I guess once they have the program in place, they also need an LLM to screen spam applications.
"We saw your profile on github and thought you might be a suitable candidate for our open position at $CRYPTO startup.
PS you must be a US-citizen, and the job is 100% on-site"
Those things seem to be blasted out with no regard for my location - I'm not looking for a developer job anyway - but certainly not one in another country.
Spamming github users seems to be the latest growth hack, and it drives me nuts. I made all my repositories archived when I started getting hit with AI-PRs to review, but I'm reaching a point where I think my life would be easier if I just closed the account.
Yeah, the playing field isn’t leveled as much as it’s simply on fire and turning into garbage. In a way it’s similar to the eternal September, but on a much broader scale.
I changed gears and moved into the video games industry at the end of 2021.
I started developing a city builder called Metropolis 1998 [1], but wanted to take the genre in new directions, building on top of what modern games have to offer:
- Watch what's happening inside buildings and design your own (optional)
- Change demand to a per-business level
- Bring the pixel art 3D render aesthetic back from the dead (e.g RollerCoaster Tycoon) [2]
I just updated my Steam page with some recent snapshots from my game. Im really happy with how the game is turning out!
> Both adults in a family will now own a car. This is required since there are not other transportation options, and sidewalks are optional.
Is this temporary or are you planning to release it like this? SimCity leaned into euclidean zoning (separate industrial/residential/commercial zones) and pocketable cars which needed no parking, and thus failed to properly showcase how ugly car-centric cities actually are. I’m sure they did it because it made for an easy gameplay loop/balancing but I’d hope we could come up with more realistic and interesting mechanics in 2026
I actually would really love that in a city planner. A game that actually simulates walkable cities versus car centric abominations and would adapt families strategies based on the availability of sidewalks, public transports and incentives.
Did you ever play cities: skylines? Keeping traffic manageable was a big part of the gameplay. Without good transit the roads would all gridlock regardless of how many lanes you add to the highways.
Played it a ton! But they stuck with euclidean zoning from SimCity and most car trips in CS don't need parking - they just disappear if there's not enough surface parking for them. They also poof away when stuck in traffic too long (unless modded).
I have been following you on twitter since I saw it. It looks amazing. Recently tried the demo. It is like under 50MB (the demo at least) which is insane these days. Placing building required construction of the building room by room which was tedious. I am sure some people will enjoy that. Will that be the core part of final game?
Thanks! Designing your own buildings is optional. The game has a feature to place zones where buildings automatically grow, but will be limited to residential and office zones at early access launch.
Can you tell more about your background? Making a sim like this also crossed my mind many times, but I learned in the past, that without much of any art skills, I would have to use resources of others or hire someone to make the graphics and so on. In the times of me playing around with RPG maker it was the missing story that was the problem. So it seems often that one core aspect is missing, when wanting to make a game. How did you learn to fill that gap, learn how to get that skilled with making the graphics?
My career background is software, but I've been a creative person my whole life.
Ive hired out help for the pixel art, and then I enhance everything with shaders (tech art).
If you're gonna make a game as an indie, you need to figure out ways to fill in your skill gaps. The competition is brutal. If you can't do it/dont have time to learn and do it, then the only other option is to hire out.
or
A lot of studios are formed from people (cofounders) who depart larger studios, so if you really want to get into the industry, you could start there and network.
I am one of those who grew up with Sim City/Transport Tycoon. I will definitely try this when it's released and go back into nostalgia but with a modern touch. Adding it to my wishlist right now. Good luck with wrapping this up towards a release!
This looks awesome! From the isometric perspective, how did you do the walls or vertical stuff in general? I have done a few game like that and always find it to be a struggle in 2D.
I'll throw a third (fourth, fifth because I know a couple of people who'd play this on Mac but who have no access to Linux or Windows) request for a Mac version on the pile.
Wow, that's comparatively very colorful. If they were kept, I wonder how HN's userbase would be different. Or how the culture would be different today. It's a different experience being able to communicate with colorful symbols. You can imply or compact a lot of words into one emoji :D
I dont use LLMs much. When I do, the experience always feels like search 2.0. Information at your fingertips. But you need to know exactly what you're looking for to get exactly what you need. The more complicated the problem, the more fractal / divergent outcomes there are. (Im forming the opinion that this is going to be the real limitations of LLMs).
I recently used copilot.com to help solve a tricky problem for me (which uses GPT 5.1):
I have an arbitrary width rectangle that needs to be broken into smaller
random width rectangles (maintaining depth) within a given min/max range.
The first solution merged the remainder (if less than min) into the last rectangle created (regardless if it exceeded the max).
So I poked the machine.
The next result used dynamic programming and generated every possible output combination. With a sufficiently large (yet small) rectangle, this is a factorial explosion and stalled the software.
So I poked the machine.
I realized this problem was essentially finding the distinct multisets of numbers that sum to some value. The next result used dynamic programming and only calculated the distinct sets (order is ignored). That way I could choose a random width from the set and then remove that value. (The LLM did not suggest this). However, even this was slow with a large enough rectangle.
So I poked my brain.
I realized I could start off with a greedy solution: Choose a random width within range, subtract from remaining width. Once remaining width is small enough, use dynamic programming. Then I had to handle the edges cases (no sets, when it's okay to break the rules.. etc)
So the LLMs are useful, but this took 2-3 hours IIRC (thinking, implementation, testing in an environment). Pretty sure I would have landed on a solution within the same time frame. Probably greedy with back tracking to force-fit the output.
I just tested this with Claude Code and Opus 4.6, with the following prompt:
"I have an arbitrary width rectangle that needs to be broken into smaller random width rectangles (maintaining depth) within a given min/max range. The solution needs to be highly performant from an algorithmic standpoint, well-tested using TDD and Red/Green testing, written in python, and not have any subtle errors."
It got the answer you ended up with (if I'm understanding you correctly) the first time in just over 2 minutes of working, and included a solid test suite examining edge cases and with input validation.
I appreciate you testing, even though it's not a great comparison:
- My feedback cycle of LLM prompting forced me to be more explicit with each call, which benefited your prompt since I gave you exactly what to look for with fewer nuances.
- Maybe GPT 5.1 is old or kneecapped for newer versions of GPT
- Maybe Opus/Claud is just a way better model :P
Please post the code!
Edit: Regarding "exactly what to look for", when solving a new problem, rarely is all the nuance available for the first iteration.
I didn't prompt anything odd, just standard prompt "etiquette", actually I significantly prompted less than I would usually do, trying to do a simple prompt like you did.
Sorry to be so blunt, but it's not surprising that you aren't able to get much value from these tools, considering you don't use them much.
Getting value from LLMs / agents is a skill like any other. If you don't practice it deliberately, you will likely be bad at it. It would be a mistake to confuse lack of personal skill for lack of tool capability. But I see people make this mistake all the time.
For games, interacting with an LLM is an improv experience. It’s a fractal that can be explored in any direction.
Video games have discrete, static goals that let a player focus on an objective. Compared to LLMs, it’s a passive experience.
People play games for all sorts of reasons (to relax, competition, to build something of their own, solve challenges).
I think this is a fundamentally different experience than what an LLM can offer.
That’s not to say LLMs can’t become a fun experience, but it’s going to take decades to develop a way to procure that experience. Look at how long it took dungeons and dragons, or any video game genre, to get to the level of polish it’s at today
It really depends on the style of game. There are gradations here. I think many designers are stuck in a static, controlling posture. Minecraft is an excellent and viral example of an alternative.
I dont think Minecraft is a fractal-like experience. The world may be infinite, but it repeats and each block that world is comprised of has set of rules that dont change. The living entities all follow their own rules. etc.
When I speak to an agent, siri, or whatnot, I am always worried that they will assume I'm done talking when I'm thinking. Sometimes I need a many-seconds pause. Even maybe a minute… For Sire and such, I want to ask something simple "Hey Siri, remind me to call dad tomorrow". Easy. But for Claude and such, I want to go on a long monolog (20s, a minute, multi-minutes).
To me, be the best solution would be semantic + keyword + silence.
I have the same issue. It gives this very weird minor sense of public speaking anxiety where I almost feel the need to write down what I'm about to say, which negates the whole purpose. Only solution I've found is using push-to-talk with some of the system wide STS applications.
I've been working on Metropolis 1998[1] for +4 years now. Custom C++ engine built with a modified version of SFML 2.5 and SQLite.
Creating my own engine was both a personal and strategic decision for me. I was really worried about running into performance issues with generalist engines, and I did not want the friction of working with someone else's mental model. Pretty sure that friction would have caused so much burnout for me. There's also the long payoff of operating in an environment that you understand top to bottom.
I ignored all the advice about making smaller games first, creating an engine first, etc. Metropolis 1998 is my first game and so far it's working out just fine. But your mileage will vary.. I started development with 10+ years of software experience and fond memories of Rollercoaster Tycoon and SimCity 2000/4.
I only add what I need. There's no level/scene editors (outside of the game being one itself :P ). No scene graphs. Shaders are coded by hand. Right now the entire game is about 45MB.
Curious how long you spent on the "engine" vs just writing the game? Based on the stuff I see on Steam you're pretty far along. I see it's Windows only which makes sense for your goals (and honestly seems very easy to run in a VM.) Was it easier to target and test for a single platform?
(Also thanks this is exactly the kind of game I'm into.)
This looks wonderful! After playing Cities Skylines 2 for the last week, all I can say is that as long as you have a half-decent traffic system, I'll be happy!
Looks fantastic! Growing up on RollerCoaster (+Transport) Tycoon and SimCity myself, this is exactly the kind of project I'd dream about building, if only I had your dedication!
reply