Show HN: AA (a.k.a. AutoApply) – open-source job-hunt tool and research project

Hi HN, I'm Nick. I'm a self-taught programmer with some B.S. in Applied Math, currently working construction but for the past few years have been building something that I think is extremely cool which is AA! My goal was to create a tool to hopefully help folks get jobs, and in the process perhaps gather and showcase job market data from a different perspective that I believe contains stories, ideas, and considerations worthy of sharing.

As of right now, AA has never successfully submitted anything, but I'm hoping that will all change very soon. Worst-case scenario users were my target audience with the goal being have it run efficiently, accurately, and reliably! I figured, if I could achieve something capable of functioning in this way then that would mean maximum availability to those who need it most.

- Customization Central Tons of ways to run: clone it, from a Flashdrive, inside Docker? Perhaps you only want it to collect qualified/aligned links or have it just apply to the links you provide or even Human in the Loop with set checkpoints? Avoid LinkedIn and Indeed and other login required sites, no problem!

- Download AA to a flash drive and plug it into a local library computer. I got the opportunity to work with my local library to offer folks sessions where they get AA up and running and while it does its thing they go and learn useful job skills and practice.

- Autonomy through Geometry: no hardcoded selectors anywhere. AA reads pages geometrically instead: Kuhn–Munkres assignment to optimally pair form labels to their inputs, convex hulls and VIPS-style segmentation for regions. Pure Python, deterministic, zero I/O, sitting behind ports.

As I mentioned, I made this to be a tool to help people find and get employed in this tough job market. Unfortunately, I knew this could only do so much which is why I tried to expand on that very idea!

- AA as a research instrument. Off by default; users can opt in, or withdraw, which deletes everything including the signing key. Zero PII collected, your profile can be encrypted with AES-256, the research data can be exported anytime, every row gets signed, and you can re-run any analysis and get byte-identical output.

- Biggest flaw to consider: once AA starts running, that device is dedicated to it until completion is reached or you end it early.

If this project caught your interest or attention there's a lot more to explore in my repo and would love for folks to go try and clone it and look around as I just pushed updates to hopefully make --install and --uninstall simple and non-destructive also use caution running session's as I was trying to impress a girl on Sunday and thought I'd try and show her my sick coding abilities and well within 6 seconds I managed to break absolutely everything which resulted in me saying saying wait every 2 minutes for 30 minutes until she left. Any who, if you are known as or even consider yourself one of the individuals below...

- Economist: What high-level things can I calculate for to have my data be more significant or applicable? What unique data could I gather that I may not consider insightful but is?

- Research Software Engineer: How bad is the damage? Any knowledge you can give me I would greatly appreciate!

- Mathematician: Sup nerd? Please look at my Math subsystem, Form Engine, Clingo code and tell me what to fix?

- Statistician: Checking out my signals_taxonomy.md would be dope and tell me the type of data I could perhaps report on that could be useful, beneficial, or applicable to current research?

- Lawyer: Would this perhaps get me in some bit of trouble?

- Any other kind of PhD person: Y'all are smart; if you're interested, ask me about my project in a way that relates to your field?

I'd love any feedback, input, suggestions, questionable insights, and funny insults that you'd like to toss my way!