[PointedResponse]

How does it work, anyway? Under the hood the infrastructure uses LLMs with very clever prompting and a ton of safeguards. It uses your uploaded resume and some info you provide at sign up (in addition to the job description) to intelligently answer whatever random questions the posting has. If it doesn't have a high confidence level in its ability to answer any of the questions, then the application will abort and notify you so that you can apply manually instead. There are also several defense mechanisms to deal with trick questions designed to catch bots. The system is designed to essentially make your applications seem like they are coming from a real human. Applications for all users are processed by a queue system that has built-in guardrails that do a few things:

Prevent applying to any single job board too many times a day to avoid being flagged as spam

Ensure each application is processed one at a time since humans don't typically apply to two jobs at once Double check that you haven't already applied to a specific position before

Spread out subscribed/zero-click applications throughout the day rather than all at once

Ensure any zero-click applications aren't generated for companies that you have already applied 3 times to in the last 3 months

If you subscribe to multiple different Saved Searches with different criteria, then they will be merged, ranked together, and the next best posting will be chosen to apply to

Take a look if you're interested and if you have any feedback you can shoot me a message in the comments, I'd love to hear it! Questions and criticism are welcome as well.

[scarface_74]

I’ve already given my opinion of the effectiveness of this. But if you did get through the HR filter and you actually described your project like you did here, there is a strong chance I would hire you.

You put some thought into it.

[tomwphillips]

I wouldn’t. It demonstrates that they haven’t thought enough to realise it worsens a systemic problem.

[scarface_74]

In the common vernacular “Don’t hate the player. Hate the game.”

He is playing the game to have a chance to exchange labor for money to support his addiction to food and shelter.

When I interview someone, I want to know whether they are the type of employee who is “smart and gets things done”. He has demonstrated both

[jitl]

This is just leveling the playing field between HR which already filters job applications with AI, and candidates who need to painstakingly research and fill out job applications one by one.

[lolinder]

Have you personally worked at a company that used algorithmic filters? Because I haven't, and plenty of people here attest that they haven't. I've only ever seen baseless speculation that it's prevalent, with no anecdotes much less data.

I'm currently participating in a hiring panel that is still working on sifting through hundreds of junk AI applications by hand to try to give the honest players a fair shot. I can certainly see the temptation to resort to algorithms now that the spam is so awful, but the causality there is reversed.

[scarface_74]

For reference, I’ve been successfully finding jobs quickly since 1996, I’m on my tenth job and I have every resume I’ve used since 2008 (leaving my second job).

I’m very qualified in my niche:

https://news.ycombinator.com/item?id=42559488

I did use ChatGPT rewrite parts of my resume this time to be more inline with the job I applied for. It didn’t make a difference since I replied to a recruiter who reached out to me based on my LinkedIn profile.

But how is it “dishonest” if I didn’t lie about my qualifications? I didn’t just take the ChatGPT output and copy and paste it. I did reword it slightly to sound like me.

[lolinder]

Using ChatGPT to require part of your resume and looking over it before submitting it to a recruiter you're already in touch with is one thing.

Mass applying for hundreds of jobs while you sleep is something entirely different that will almost certainly lead to lying on the submitted resumes.

[rgbrgb]

Well if that last step is what made it feel honest, many of these tools are basically a way to bypass that. I think it crosses into dishonesty when you include an AI cover letter that says you want the job/would be good at it but you haven’t even looked at the post or evaluated the company. For example if you don’t actually want the job (have seen this).

[mattfrommars]

Are you using any kind of job scheduler to do the job for you in the background? Are you using Apache Arrow for job scheduling since you mentioned you need to apply to jobs behind the scene and spread out across day and night.

[PointedResponse]

Conveniently, I decided early on that I would rate limit applications based on the job board, so that any single user doesn't have too many applications sent to a single job board site like greenhouse or lever in one day. So for now, to spread them out I have an ec2 querying applications off the application postgres table every 15 minutes. It takes the next best job posting per board for each subscribed user and puts it into a queue to be processed. 1 application per user, every 15 minutes, with 1440 minutes in a day, means ~96 applications per user to any given job board per day. The applications are then processed off the queue by another ec2 synchronously to make sure only one posting is applied to at a time per user. In addition to the 96 application cap that comes naturally with this method, there is also a hard cap of 150 applications per day per board so that if the user is also going crazy using the "One-Click" applies, those in addition to the fully automated / "Zero-Click" applies won't get them flagged for spam by the board site.

[dennis_jeeves2]

I presume you will have to simulate a user filling in the application on the browser, how is that part done? given all the anti spamming measures.