The Infinite Loop: Why the AI-Hiring Arms Race Is Producing Worse Candidates, Not Better Hires

Somewhere in the last eighteen months, the average job application got dramatically better and the average job candidate got quietly worse. Both things happened at once, and they are not a coincidence. They are the same event, described from opposite ends of the same pipe.

The packaging improved because a model writes a cleaner bullet point than most humans manage on a Tuesday night. The candidate degraded because describing your own competence — the thinking, the choosing, the cutting — is the practice that makes competence visible to you in the first place. Outsource it and you don't just lose the words. You lose the rehearsal.

This matters more than any other story in hiring right now, because hiring is the gate to income and income is the gate to everything else. When the gate stops measuring anything real, the bill lands hardest on the people who believed the system was fair and followed the instructions exactly.

My argument is this: the AI-hiring arms race is not failing to improve hiring — it is actively eroding candidate quality while hiding the decline behind better formatting. And the damage is being laundered through a two-speed market where human attention is quietly becoming a premium good. That's the failure nobody is pricing in, and I'll defend it to the last word.

The Machine Is Now Grading Its Own Homework

Follow one application through the modern pipeline. The job description was written by a model trained on older job descriptions, so it arrived inflated with requirements no team has ever actually needed. The candidate pastes that posting into a tool that rewrites their resume in the posting's own vocabulary. The applicant tracking system then scores the resume against the keywords it just copied. Finally, the recruiter's summarizer reads three hundred of these and hands back a dozen "top matches" in confident paragraph form.

Nobody in that chain is lying. Nobody is thinking, either. It works like a photocopier copying a photocopy: every pass loses a little specificity, gains a little generic confidence, and by the seventh generation you're holding a page that looks perfectly crisp and says absolutely nothing. The machine is grading its own homework and awarding itself a B-plus.

This is why the quality question has become so hard to answer honestly. Recruiters say candidates are worse. Candidates say interviews are worse. Both are accurately describing the same degradation from opposite ends of a lossy channel — and neither side has the vocabulary to name what happened, so they blame each other.

Legibility Is Not Competence

The filter was never measuring skill. It was measuring how closely your language matched the language of a posting that was itself generated from previous postings. That is a test of fluency in the industry's dialect, not a test of whether you can do the job.

In effect, we exported the worst idea in education — teaching to the test — into the most consequential decision in adult life. And it behaves exactly the way it always does: the person who studied the rubric beats the person who did the work.

Consider two applicants. The first rebuilt an intake process nobody wanted to touch, cut a five-day delay down to one, and has no idea how to phrase it. The second read the same forty listicles everyone read this month and wrote that they "spearheaded cross-functional synergy to drive operational excellence." Guess which one clears the threshold. Guess which one the hiring manager actually needed on the team by Friday.

That is the decline in candidate quality nobody wants to say out loud. It isn't that people got worse at their jobs. It's that the artifact stopped correlating with the work — and the industry responded by tuning the artifact harder instead of asking what it was for.

The Atrophy Nobody Is Measuring

Skills decay when you stop using them, and the arms race has quietly removed the daily use.

Think about how you navigate a city you've lived in for years versus one you only ever reach by GPS directions. The app doesn't just replace your sense of direction — it prevents you from building one. Careers work identically. What used to be an hour of genuinely hard thinking about your own record — what did I actually change, what would have broken without me, where am I weak — now takes four seconds and a prompt.

The bill comes due in the interview, the first unscripted moment in the entire process. The candidate cannot walk through their own bullet point. Not because they lied, necessarily, but because they never did the thinking that makes a story yours to tell. The hiring manager hears vagueness and concludes deception. The candidate feels exposed and concludes the process is rigged. Both are looking at atrophy and diagnosing character.

That's the cruelty of the loop. It hides the decline until the exact moment the decline is most expensive, then charges the candidate for the discovery.

The Two-Speed Market: When Human Attention Becomes a Luxury Good

Watch hiring in real time and you'll see it split into two lanes that barely acknowledge each other.

Lane one is algorithmic: infinite volume, instant rejection, almost no hires relative to the effort it consumes. Lane two is human: a named introduction, a referral, a note that references a specific problem a specific team is currently wrestling with. Slower. Smaller. Gated. Dramatically more effective.

The uncomfortable part isn't that lane two exists. It's that lane two used to be the default. Human connection was how hiring worked — badly, unevenly, and often unfairly to outsiders — but it was how it worked. Now it's a premium feature. The thing that actually decides outcomes has been repriced as a luxury good, available mostly to people who already have the network, the alumni channel, the former manager who remembers them fondly.

That's the real two-speed market: not fast versus slow, but seen versus unseen. And it should make you angry, because the only lane left open to a stranger with no network is the one engineered to filter them out.

Breaking out of it does not require a bigger engine. It requires fewer, better, human-shaped moves: a target list of actual names, a narrative you can defend out loud without notes, and a follow-up rhythm that a person on the other end would recognize as respect rather than automation.

That is the entire design principle behind Job Search Pass. Access comes as a single pass — 90 or 180 days, one payment, no auto-renew — because the purpose of a job search tool is to get you out of the tool and into conversations. The software handles the compression: sharpen the story, map the names, keep the cadence honest, apply to fewer things with more precision. Then it steps aside. A subscription that renews while you circle a queue is not support. It's a meter running on your anxiety.

Stop Auditioning for a Machine That Can't See You

The loop will keep turning for exactly as long as people keep paying to be inside it. Every tool bought to out-optimize another tool is another quarter-turn of a crank that goes nowhere.

You have a way out that costs less and works better: choose signal over volume, named humans over portals, rehearsal over formatting, a finite pass over an infinite fee. Machines are extraordinary at producing volume and hopeless at producing trust — and trust is still the only currency hiring runs on.

So stop auditioning for a machine that was never able to cast you. Buy the pass. Do the work. Go be seen by a person.