JobsByReferral.com and an eventful 24 hours on the front page of Hacker News
Throughout June 2025, several of my developer friends had been laid off.
We were discussing the state of the job market in tech (which is... Not Great, Bob), and were collectively surprised to find that searching for jobs on LinkedIn no longer had an "In My Network" filter (but more on this later -- it's still there, sorta).
The conversation kept coming back to how frustrating the job search had become in general. Cold applying to jobs has become ineffective when you're competing with hundreds or thousands of applicants per job post. AI tools make it easy to tailor resumes and cover letters to each position, so companies are drowning in applications.
In my 20 years in tech, the majority of roles I've landed were through a referral of some sort. Knowing what roles are available is the first step in that process -- you can't leverage your network if you don't know where the opportunities are.
I was quickly nerd-sniped and spent a couple of hours with Claude Code vibe-coding an app that would do something simple: take your personal LinkedIn connections data and show you open jobs at companies they worked at: JobsByReferral.com.
The 15-Minute MVP
Within 15 minutes I had a working approach: you could request your own personal LinkedIn data archive which contains a Connections.csv file. This has your connections' names, company names, and email addresses. From there the app I created would analyze your network and show you the companies you're connected to (and who works there):
Then, the app runs each company through JSearch, which uses the Google for Jobs API:
I shared this with my friends who were job hunting, and some of them found it pretty useful. They were finding companies they didn't realize they were connected to, and it gave them a systematic way to approach their network rather than just pouring through job boards.
Tech Stack
I built this with Next.js on Vercel, and used jszip to do the data archive ZIP extraction entirely on the client. This was one of my core tenets of this project: don't let personal data touch a server. As I had noted in this HN comment:
we do not pass user data to the backend endpoints, which you can verify by viewing the network calls. When you upload the ZIP or CSV the extraction/parsing happens entirely client-side, and then we use auto-generated IDs to map connection data from the JSearch API response to the client-side stored connection data.
Searching for "In My Network" jobs on LinkedIn
As I had noted above, turns out we were somewhat mistaken on the lack of an "In My Network" feature in LinkedIn Jobs search. They're rolling out a new AI-powered Jobs Search platform, so the previous filter for "In My Network" is built into the query itself instead of being a toggle you could flip, like the old search.
So the functionality is still there but it's less discoverable, which is what motivated this little experiment in the first place. I think there's an interesting sidebar here about how AI/LLM integrations in products are hiding advanced features that many folks rely on and expect from tools, but that's another post!
24 hours on the front page of Hacker News
I threw it on Hacker News as a Show HN post and rushed out the door to take my kid to dance class. A couple hours later I noticed 50+ realtime users on my Fathom dashboard, and knew it had popped to the homepage - where it sat for about 24 hours!
This was a fun project to build and was neat to see some traction, though the popularity is likely driven by the state of the job market, which is more than a little bittersweet.