I Applied to 847 Jobs. Here's What I Learned About Being Invisible.

I. The machine

The system is not broken in the way they say

Nobody says this part out loud at recruiting conferences: the ATS is not broken. It's working exactly as it was built to work.

It wasn't built to find the best candidate. It was built to take 400 applications and turn them into a shortlist — fast, defensible, cheap. Reducing a pool and identifying the best person are not the same job, even though we've been pretending they are for twenty years.

You've probably seen the stat: ATS software auto-rejects 75% of résumés before a human ever opens them. It gets repeated so often it sounds official. But the number doesn't have a reliable published research source behind it, although in reality the share filtered out is still quite high.

The real problem is more complicated — and better documented. A 2021 Harvard Business School and Accenture study found that more than 90% of surveyed employers used recruiting systems to filter or rank candidates. And 88% of employers agreed that qualified high-skilled candidates were screened out because they didn't match the exact criteria in the job description. Hidden Workers: Untapped Talent.

What most systems do isn't simply delete résumés. They filter, rank and narrow the field. A CV that doesn't match the right criteria doesn't necessarily get rejected by a robot. Sometimes, it just sinks.

And when a recruiter does open it? In a 2018 eye-tracking study by Ladders, recruiters spent an average of 7.4 seconds on an initial résumé screening pass. Long enough to scan for a shape. Not long enough to read for depth.

The architecture migration that saved the company millions. The mentorship program you built from nothing. The call you made that stopped a cascading production failure at 3am. All of it sits under a bullet that gets skimmed, not read.

II. Optimisation

Why optimisation can't fix this

So you try harder. Templates. ATS simulators. STAR-format bullets. Metrics added, then removed. One page, then two. Premium LinkedIn. "Open to Work," switched on.

Still invisible.

Optimisation assumes the system is neutral — that if you just find the right words, you get through. It isn't neutral. It was built to filter you out, not read you in.

There's an older problem underneath the new one. The résumé is roughly 500 years old. Leonardo da Vinci wrote something close to the first one in 1482 — a list of what he could do, sent to a stranger he'd never met, hoping it would speak for him.

The format has changed since then. Fonts, PDFs, keyword scanners, applicant portals. The logic hasn't. You still send a document into a void and hope someone reads it as a person instead of a shape.

Putting a smarter algorithm on top of that model doesn't fix it. It just produces faster rejection. Faster noise, not better hires.

III. A different logic

The structural alternative

So we built something with different logic, not a faster version of the same one.

You upload your CV once. Maybe a LinkedIn export if there are gaps. That's it. No forms asking you to retype every job you've ever had.

We turn it into an evidence-backed profile — a vairee-card. Every claim tied to a source: your CV, your LinkedIn, your GitHub, your Stack Overflow. Say you contributed to an open-source project — we check. Say you're fluent in Python — we look for the evidence, not just the word.

Then we write what we call a recruiter's brief. Not your bullet points. How a headhunter would actually describe you to a hiring manager.

"Jan is a Staff Backend Engineer with a track record of owning high-availability payment infrastructure. His last role included a six-month migration that rebuilt the core transaction layer and cut cascading-failure risk — a project that lived under 'maintained AWS infrastructure' on his old CV. Four stack generations, seven years, zero major production incidents."

Not a keyword match. A signal.

You approve it. Edit it. Or skip it. Nothing goes live without you.

Then you get a share link, not a PDF. No raw CV in fifty recruiter inboxes. No salary exposed on first contact. No login required for someone to view it. You update it once and it stays current. You decide what's visible, and to whom, until you say otherwise.

IV. Where we are

The important part

We'll be straight about where this stands. There's no algorithm quietly shopping your profile to companies overnight. No phantom AI running in the background while you sleep.

What exists right now: you build your vairee-card. If you switch it on (activate vairee-talent card), our team — actual headhunters, the kind who've spent twenty years doing this the hard way — reads it and reaches out to companies directly, when there's a real fit.

That's not a lesser version of automation. A human vouching for you beats an algorithm that doesn't exist yet.

This is MVP. We're building it in the open, with real people using it, which means it gets better because you tell us what's broken.

The best people are not hard to find. The system was just never asked to find them — it was asked to filter, and it does that well.

That's a design choice, not a law of physics. It can be redesigned.

V. Appendix

Sources