Synopsis
This post describes my first two months of job search, particularly the challenges faced when relying on Linkedin, and an alternative low-cost (but medium effort) to detect and respond rapidly to relevant job postings.
Finding a job
2026 came with its lot of change, and I decided to leave my previous company and search for a new adventure. LinkedIn has become for many the default place to find a job, and is therefore where I began my search. I hadn’t actively looked for a role since 2019, having over the past years evolved within the company I worked for. My first look was positive: lots of jobs to be found, and lots posted recently too. LinkedIn’s personalized recommendations were 90% completely useless and 10% surprisingly interesting, so I ignored them and used keywords to search.
I realized that “optimization” was taking place, and my search was “influenced” by companies paying to promote their open roles: Most of the jobs I’d find were broadly the right kind of work but didn’t actually contain the keyword I was looking for. It therefore took effort to look through the details and find positions that would seriously consider me.
As days passed and too few emails came back (42% of the 17 applications I sent in April were still without response in mid-May), I realized that most of the jobs “Posted in the past 24 hours” were actually “Reposted” in the past 24 hours; a mountain of zombie jobs I would see week after week, always with dozens or more applicants within hours of (re-)posting, and while I still awaited their response to an application sent weeks prior. Looking at further pages of the search wasn’t beneficial, as I found worse recommendations by the page, until every single role came from a recruiting firm or Mercor (a company hiring contractors to train AI).
Frustrated, I decided to create a better job platform: fortunately common sense got the best of me early, and I looked instead at the million or so alternatives that already exist. I read about Welcome to the jungle and Omnijobs: the former had too few positions in my country (the Netherlands) while the latter seemed to fail my keywords search worse than LinkedIn.
More frustrated, I started looking at the job portal of companies that interested me, and found many positions recently posted that had never surfaced across my searches. I applied.
Being quick to find a fitting job
Responses started to come back: as wonderful as my profile is, every company had received incredibly fitting candidates and wished me the best in my search. I doubled down, with limited results. It took a few applications through referral coming back the same way to realize that something else was at play. According to my referrers, recruiters received hundreds of CVs and often kept the first N of the pile. Since timing was so critical, I spent a lot of my days refreshing LinkedIn job searches and company portals, looking for a fitting job and applying as rapidly as I could: finally, a few interviews came through.
As days passed and interviews took place, I decided I’d automate as much of this painful process as possible to save time, avoid going insane, and hopefully land a job. After all, how hard can web scraping be in 2026?
It turns out that scraping has become significantly harder than 10 years ago, and you now have to pay a SaaS to scrape for you because companies got wiser and fight scraping better. This was not a pleasant discovery, but I had time and thought it could make a nice side project. I chose n8n to orchestrate the process, as one of the jobs I was interested in asked for experience with it, and began mapping the process to automate:
- Scrape the job portal of a company I’m interested in
- Keep only recent openings
- Evaluate how well these roles fit my background and aspirations
- Alert me when a high-fitting job is opened, so I can apply right away
- (Note: I considered automating the application too but felt it would add significant complexity and could result in many missed/imperfect applications)
The specifics of how this is done in n8n is not particularly groundbreaking, and I will therefore simply link the workflow in appendix below for anyone interested in replicating it on their own account (see Appendix A). In a nutshell, we use Apify to call an API that specializes in scraping job portals, and filter for specific companies, full time positions, and jobs freshness. We then pass this list of jobs to Claude (via an API call), alongside a prompt explaining the user’s experience and job aspirations, and ask to receive back a score reflecting how well the job matches. Finally, we email ourselves the jobs with high scores alongside a quick summary justification and the link to apply.
Avoiding costs and still finding a fitting job
Unfortunately there is no free lunch and after a few days, if you are interested by many companies, or by companies opening lots of roles, Apify will raise alarms about their free tier being exceeded and a 29$ per month subscription becoming required to continue. While you could pay, you can also create new email aliases if you have your own domain, or simply use temporary email providers to create a new account for the next days or week (note: this is blocked as of mid-May).
A few more days, and n8n will tell you that time passed and you should now pay them 20€ per month for their subscription. The good news is that the solution above works just as well, although I specifically recommend here to use aliases instead of temporary emails, to keep security tighter since your n8n account contains credentials to various APIs.
Finally, there is a temptation to ask more from Claude: prepare my CV, prepare a motivation letter, etc. Trial and experimentation will quickly help you figure out that API costs increase rapidly as your documents are passed back and forth through the calls, alongside a long prompt and longer job description.
Additional considerations
- The workflow does not test whether all companies of interest are actually scraped, and some might well use a system that isn’t compatible with Apify’s scraper.
- Claude’s scoring is also something that requires attention and indirect fine tuning through prompt edits.
Appendix A. Job search workflow
The .json file in the repository above can be imported into n8n and used rapidly. You will need to update all authentications to use your own API keys. You can also replace Claude by ChatGPT, SMTP email by Gmail, etc.