LLMs are winning the job application arms race
With a little scripting ability, it is getting easier to use LLMs to spam job applications.
The developer job market is tougher than it was during the pandemic. There were wide-scale tech layoffs over the past 2 years, which coincided with fewer open jobs. To make it even worse, it’s getting easier to flood job applications with LLM-generated slop.
I personally felt the tough hiring climate. When I was job hunting at the beginning of 2023, FAANG companies were undergoing massive layoffs along with the rest of the industry. They laid off thousands of employees, and surely many of them wanted jobs in tech. I was now in competition with them.
I knew what I wanted from a new position:
I wanted to have a lot of users.
I wanted to work on the company’s core business.
I wanted to be proud of the business model.
This means that I was limiting the companies that I wanted to work for. This put me at a huge disadvantage! Obviously, the best strategy for landing a job is maximizing the number of jobs you apply to. It becomes a numbers game: the most outbound outreach leads to a lot of interviews leads to a few offers leads to a negotiation leads to a maximized salary. If I actually dare to have preferences, this puts me at a huge disadvantage. Anyone doing the “optimal” thing will dilute my applications to the jobs I prefer even if they weren’t explicitly interested.
And this is truly happening. If you’ve spoken to a hiring manager recently, they’ll tell you “I had all the applications I need within a day of posting the listing.” Job seekers do a full-court press when they’re applying for jobs these days. When Etsy — my former employer — had a round of layoffs last year, over a dozen people asked for referrals to various open roles at my current job. Everyone knows how the game is played and they’ve optimized for it.
But when I was interviewing, I still had one ace up my sleeve: if you wanted to have a strong job application, you needed to do most of the work yourself. LLMs in January 2023 were impressive, but far from perfect. Since then, models have dramatically improved (ChatGPT 4, Claude Sonnet, etc), and the price of LLM calls keeps dropping. 10 years ago it would have been insane to say, “I want to apply to every open software engineering job in the world,” but nowadays it’s easy to imagine that it’ll happen soon.
And in fact, someone basically built a bot to do that.
“How I Automated My Job Application Process. (Part 1)”
So I did what any sane developer would do - I built a system to automate the whole damn thing. By the end, I had sent out 250 job applications in 20 minutes. (The irony? I got a job offer before I even finished building it. More on that later.)
Let me walk you through how I did it.
If you read through the post and its sequel, there is no real magic. He copy/pastes the HTML into his program. A LLM converts that HTML into structured JSON. He would then generate a cover letter email that was completely personalized based on information that it found in the job posting, using relevant experience of his to explicitly tailor his resume to the job posting.
The secret to good cover letters? Context. I fed my resume into the LLM along with the job details. This way, the AI could match my experience with their requirements. Suddenly, those "I'm excited about this opportunity" letters actually had substance.
[…]
The prompt does a few clever things:
Forces structured output - no wishy-washy responses
Tracks which of your skills match the job requirements
Identifies any missing info that could strengthen the application
Generates both HTML and plain text versions (because some job portals hate formatting)
And here's the kicker — it fails fast if critical info is missing. No more generic "I saw your job posting" emails. Either the cover letter has substance, or it doesn't get sent. Period.
(I start all my prompts with ‘please’, so that when AI eventually takes over, they would consider me friendly 😁)
I wondered if he was the first person to come up with this idea, but obviously he was not. It’s not hard to find blog posts where people explain how they are using LLMs to supercharge parts of the job application process, like finding jobs that look for your skills, or automatically generating resumes or cover letters. In fact, if you want you can even pay for services that do this. Hell, if you want to cheat on your interview, people have your back.
A quick search on Hacker News over the past month reveals that it’s pretty easy to apply to jobs “while you sleep” by paying to apply to jobs on a specific job board.
Before slop-pocalypse — where we all choke to death on LLM-geneated slop — took effect, the hiring space was filled with “outbound slop,” i.e. meaningless content from companies to potential applicants. Here’s what I mean: Recruiters just set up cold outreach campaigns for developers and they got spammed with an endless sea of emails. For some unknown reason1, my mother’s email address gets occasional cold email outreach for jobs for me. I had to explain to her, “There aren’t real people on the other end. You don’t need to email them. They’re not waiting for a response. You don’t need to forward them to me.” And this was true! It’s all automated campaigns. They send out emails that sound inviting, and when you ignore them the automated systems send out a few ticklers.
But now, the hiring space is filled with “inbound slop,” i.e. slop from candidates to companies. Any post will get an infinite amount of interest, and the hiring manager needs to find effective ways to filter it down.
With the collapse of traditional online hiringk, where is all of this heading? Who knows, maybe there will be alternative hiring approaches that are automated versions of traditional headhunters, where they list companies where you’d want to find people with X job (“if you need someone on your recommendations team, then you should try to poach recommendations employees from these 30 companies”). Maybe they will just point an LLM at the slop and blindly accept whoever comes out the other end. And hopefully, some brilliant person will crack the code and get their payday. If they can find a way to efficiently match employees with companies, they earned it.
And I truly hope this was an honest mistake!