The guy who automated his way out of unemployment just open-sourced the playbook—and the fact that being honest about using AI got him *more* interviews says everything about where hiring is headed.
The Summary
- A geophysicist turned AI engineer open-sourced the exact framework he used to land his next job: 69 applications, 20 interviews, 1 signed contract as an AI engineer
- The workflow runs locally on Claude Code: scrape jobs, evaluate fit, tailor CVs, write cover letters, prep for interviews—no SaaS middleman, no data upload
- Counterintuitive result: telling employers he used AI to apply *started conversations* instead of disqualifying him
The Signal
Mads Lorentzen got laid off in late 2025. Geophysicist by training, no AI engineering background. Six months later he started as an AI engineer. The bridge between those two facts is a GitHub repo with 300 stars and climbing.
The framework is straightforward: fork the repo, fill in your profile (skills, experience, preferences), point it at job boards, and let Claude Code run a three-stage pipeline. First it scrapes and evaluates postings against your profile. Then it enters a drafter-reviewer loop—one Claude instance writes the application materials, another reviews and refines them. Finally, it preps you for interviews based on the role and company research.
"Sixty-nine tailored applications, twenty first interviews, and one signed contract later, I started as an AI engineer in June 2026."
The technical choices reveal someone who's thought past the demo. It runs entirely local. No API calls to a startup that might pivot or shut down. No uploading your resume to someone else's database. You own the infrastructure, you own the output, you own the data. The core workflow is language- and country-agnostic. The job portal scrapers are built for Denmark (Jobindex specifically), but the architecture is designed to be forked and adapted.
What matters more than the code is what happened when Lorentzen used it. He didn't hide the AI assistance. He told every employer upfront. And instead of being a liability, it became a signal. It demonstrated technical fluency. It sparked conversations about automation, tooling, how work actually gets done now. Twenty companies gave him a first interview. One hired him for a role he had no formal credentials for.
The shift isn't that AI can write cover letters—of course it can. The shift is that *using AI to write cover letters is becoming a technical skill employers want to talk about*, not a red flag they screen for. Hiring managers who understand what's happening see someone who can scope a problem, build a workflow, and ship a result. That's the job description for half the roles being created right now.
Key dynamics at work:
- The drafter-reviewer architecture (one agent writes, another critiques) is the same pattern showing up in code generation, legal doc review, and content production
- Local execution matters: this runs without API dependency, which means it works in five years even if the startup ecosystem around AI job tools has churned completely
- The "honesty as differentiator" angle only works because most candidates still aren't doing this—early-mover advantage has a shelf life
This is what the agent economy looks like at the individual level. Not some far-future scenario where AGI does your job. A laid-off geophysicist who built a tool that let him punch above his weight class in a hiring market that wasn't designed for career pivots. He didn't use AI to fake expertise. He used it to get his actual expertise in front of people who could use it, faster and more precisely than he could have done manually.
The Implication
If you're hiring, the candidates using tools like this are the ones who understand how work is changing. If you're job hunting, the question isn't whether to use AI in your search—it's whether you can explain *how* you're using it in a way that demonstrates capability instead of desperation. Fork the repo. Adapt it to your market. Own the infrastructure. And when you get the interview, don't hide the process. Make it part of the story.
The marginal cost of a tailored application just dropped to near zero. The differentiator is no longer effort—it's taste, judgment, and the ability to deploy the right tool for the problem.