This morning, while reviewing our Senior Full-Stack Engineer inbox at Heavy Chain, an uncanny sequence arrived.
At 10:01 UTC, an application landed from "Peter Addo."
At 11:07 UTC—sixty-six minutes later—another arrived from "Henry Kareta."
At 12:18 UTC—seventy-one minutes after that—a third landed from "Robin Johansson."
Three applications in three hours. On the surface, that looks like a healthy inbound pipeline. But look even slightly closer, and the seams immediately give way.
All three followed an identical handle format: [firstname][lastname][3-4 digits]@[gmail|hotmail].com. Clean first names, clean last names, no middle initials. Peter and Henry closed their notes with the exact same geographical sign-off: "I am based in Tallinn, Estonia, available immediately, and seeking a remote position." Robin omitted Estonia from his prose, but his email arrived stamped with the exact same +0300 Eastern European Summer Time header.
Then there was the language.
Peter claimed "10 years of experience." Henry claimed "more than 11 years." Robin claimed "more than a decade." All three spoke in the uncanny cadence of a single LLM prompt straining to sound like three distinct humans:
"I have built and maintained full-stack platforms, APIs, and customer-facing products across fintech, payments, insurance, and SaaS environments..."
"In my current role... I develop Python services and TypeScript/React product features... extending PostgreSQL data models... brownfield modernization... practical AI capabilities rather than treating AI as a separate experiment."
The phrasing echoed across all three letters like a chorus: brownfield modernization, relational data models, customer-facing products, retrieval workflows. The attached resumes were styled differently, but beneath the formatting lay permutations of the exact same semantic fingerprint, tuned directly against the keywords of our posting.
And, of course, the seventy-minute timer. Whatever script was driving the loop knew enough to space out submissions to slip under crude spam rate limits, but lacked the sophistication to simulate genuine human entropy.
It was clockwork.
THE GENERATION ASYMMETRY
It is tempting to dismiss this as an isolated curiosity—or perhaps a clumsy attempt by an offshore proxy shop, or worse, a DPRK remote-worker syndicate hunting for another crypto/fintech payroll to tap. Those are real security dynamics; every engineering leader in tech is now having to learn threat modeling simply to review resumes.
Yet the broader observation is economic.
For the last thirty years of tech hiring, the inbound application was governed by a crude form of proof-of-work. Applying for a job carried real marginal cost. Even a candidate blasting out generic resumes had to spend time: finding the link, uploading the PDF, typing into an ATS, tailoring a sentence or two. That human friction was a natural throttle. It kept the ratio of applications to open roles within a manageable order of magnitude.
AI has reduced the marginal cost of applying to zero. Effectively zero.
An automated agent can monitor job feeds across the web, parse a job description, generate an impeccably tailored cover letter, hallucinate or spin a plausible resume with 10–11 years of experience in the exact stack requested, spin up a throwaway email account, slap together a fresh LinkedIn profile with a numerical hash tail, and dispatch it on a polite one-hour jitter.
Total cost to the applicant: a fraction of a cent in inference tokens.
Time spent by the applicant: zero seconds.
"When generation costs $0.001 and verification costs $15.00 of human attention, the system does what any asymmetric economic system does: it floods."
The problem, of course, is that verification does not scale at inference speeds.
Reading an email, opening a PDF, evaluating a GitHub profile, verifying references, and assessing credibility still demands genuine human cognitive bandwidth. Even a cursory initial scan burns two to five minutes of an engineering leader's day.
When generation costs $0.001 and verification costs $15.00 of human attention, the system does what any asymmetric economic system does: it floods.
THE DEATH OF THE OPEN FRONT DOOR
In engineering, we talk frequently about the "verification problem" in software delivery—the reality that an LLM can write 2,000 lines of code in seconds, but a human still has to understand, test, and maintain it.
We are now watching the exact same dynamic play out across hiring.
When inbound signal collapses into pure noise, the rational response from companies is not to build better filters. Filters are just another boundary that an adversarial LLM will learn to game by next Tuesday. The rational response is to close the front door entirely.
You can already see this retreat happening across the industry:
- The Retreat to High-Trust Networks: Warm referrals, ex-colleague backchannels, and closed invite-only talent rings are becoming the only channels that senior leaders genuinely trust.
- Live Proof-of-Work: Demanding un-gameable friction early in the funnel—a short, unscripted async video, a hyper-local conversational probe, or an interactive pair-programming session where an AI prompter cannot hide behind a proxy speaker.
- Outbound Over Inbound: Teams increasingly ignore public inbound queues altogether, preferring to proactively hunt candidates whose work leaves a verifiable public artifact trail: open-source commits, technical writing, and provable production footprints.
THE INVISIBLE CASUALTY
The irony is that companies are not the primary casualties of this shift. We simply configure tighter spam rules and lean harder on our personal networks.
The real victims are legitimate candidates.
"The open front door of the internet is getting locked—not because companies do not want to hire, but because the cost of listening has finally exceeded the value of the noise."
The junior engineer trying to break in, the brilliant self-taught developer without an elite network, the quiet senior engineer who actually spent eleven years building reliable PostgreSQL pipelines in Tallinn without blasting three identical cover letters to a stranger's inbox—they are the ones drowned out in the deluge. Their thoughtful notes sit unread beneath seventy variations of the same prompt-engineered hallucination.
When everyone can generate infinite signal, signal becomes indistinguishable from noise.
For now, Peter, Henry, and Robin will quietly go into the trash bin. But their arrival on a Friday morning, seventy minutes apart, is a quiet indicator of where we are. The open front door of the internet is getting locked—not because companies do not want to hire, but because the cost of listening has finally exceeded the value of the noise.
AI attribution: 3/4 — Human-originated, AI-shaped. What this means
Jason Vertrees is the founder of Heavy Chain Engineering, which helps lower middle-market vertical SaaS companies and PE firms turn scattered AI usage into measurable delivery leverage — 85% faster feature velocity, six-to-eight-week projects shipped in days. If you want help building an AI-native engineering organization, book an AI Delivery Assessment or email jason.vertrees@gmail.com.