Breaking My Own Rule (Again): AI and the Job Hunt
As I look for work, please support me with a donation to the project development I'm spending in the Atmosphere — donate hereI've spent a good amount of (virtual) ink on this site being suspicious of
generative AI and arguing that protesting it is too easy when the
actual power sits with capital allotment towards its development and research,
not with the model weights. I stand by that. I also admitted a few weeks
ago that I broke my own stance and started using it to follow the industry,
in hopes of improving my chances at getting a job at all. It's no different
than me returning to use GitHub for all of my project development, despite also
using things like https://tangled.org to begin hosting my code (as confusing
as it can be at times) and exploring non-Git tools like (not but sponsored
heavily by) Google's jj.
The tools
Jack and Jill has been doing some of the job-discovery legwork lately — surfacing roles I wouldn't have found by manually refreshing the same three job boards every morning, and helping tailor cover letters per role instead of me copy-pasting the same paragraph with the company name swapped out[1]. I won't pretend the cover letters it produces are precious; they're serviceable, which is exactly what a cover letter needs to be when you're sending out five a day. I also am on LinkedIn a lot - I yap about it and the bugs I find (which were, ironically a result of uBlock Origin) and pipe roles into Jack to help keep a second place to track all of my applications (I keep a separate spreadsheet on my computer for my own quick math).
The other half of this is local: I use things like OpenCode, Pi and Claude Code against my own resume to tailor it per role, when necessary. Since my resume's yet another template that lives in this site's repo; that makes the act of tweaking things very quick and cheap. Having an LLM read a job posting and my actual work history side by side and tell me which skills to foreground (or which experience bullet is dead weight for this particular application) is a much better use of my time. I was doing this manually and was only able to get about 6 applications a day in when doing longer reviews. Now it's at least three times that.
"Damn, son, now AI helping you look for work?"
None of this resolves the argument I made in the ease-of-protest piece: the workers who build these tools still work for the people I have every reason to distrust, and the fact that a consumer-facing product happened to be useful to me this month doesn't make the incentives underneath it any better. Like I mentioned above, this is mainly about increasing the number of shots I can take in an industry that's already invested heavily into these tools. What it does mean is that under enough precarity, the distance between "principled abstention" and "can't afford to lose another week" collapses fast.
I don't know yet if any of this moves the needle in a market where referrals matter more than the work itself. But it's cut down the time between "found a posting" and "applied" enough that I can throw a wider net without burning out on the process before lunch.
This is obviously specific to me but I do have a "gold tier" version of my resume that's a collection of all of my work notes that I still have and I whittle it down to bullet points that highlight what important skills are needed. This is something that only really works in digital tooling because updating the main resume (which is outside of the repository) is only useful for my personal note taking and would have been too verbose for a normal resume. ↩︎