The deluded dimness of the modern job hunt

AI-assisted founder authorship

AI-assisted author identity used at John N. Farmer’s request. John is OMIKINA’s founder and accountable publisher, not a fictional reporter. Text is prepared with AI assistance; authorization to publish does not itself establish completed human editorial review.

Key points

• The audit records 547 application actions for distinct roles between September 27 and October 12, 2026; this does not prove every employer received or reviewed the materials.
• Employer replies received September 27 through October 14 contain at least 98 distinct rejection cases from 79 companies: 80 match exact application-ledger entries and 18 remain unassigned.
• Silence remains unresolved, duplicate notices were consolidated conservatively, and these figures are not a rejection rate for the 547 actions.
• AI-assisted applicant and employer workflows still need meaningful human judgment; custom job-specific decks remain an unproven experiment.

Opinion

The modern job hunt is exceptionally good at making you feel busy.

Update the profile. Rewrite the résumé. Reorder the portfolio. Explain the same experience in a slightly different box. Make the cover letter sound enthusiastic without sounding desperate. Submit. Repeat.

With AI, I made that routine faster. We built a process around verified experience, updated professional materials, relevant portfolio links and responses tailored to each posting. Agents helped research openings and prepare applications. Receipts gave us a way to distinguish work completed from work merely attempted.

In my earlier article about that process, I wrote about the possibilities. This is the less flattering follow-up: efficiency can become a very sophisticated way of standing still.

Sources: S1

My rough impression was that we had sent 600 applications. The audit supports 547 recorded application actions for distinct roles between September 27 and October 12: 511 portal confirmations or acknowledgements and 36 roles addressed through recruiting emails. One additional email bounced and was removed from that total. These are recorded actions, not proof that every employer received or reviewed my materials. The search includes creative leadership and technical AI roles.

A review of employer replies received from September 27 through October 14 found at least 98 distinct rejection cases from 79 companies. Eighty match exact entries in our application ledger. Eighteen remain unassigned: some have broad or missing titles, while others name roles absent from the ledger. Duplicate notices were consolidated conservatively. This is a verified minimum of refusals received during that period, rather than a rejection rate for the 547 recorded actions. Silence remains unresolved.

That distinction matters. I am angry enough about this experience without improving the story by inflating the numbers.

What wears me down is the shortage of useful information. An email can say the application was carefully considered and then give me almost nothing I can use. Was the experience insufficient? Did the role need a narrower specialty? Was compensation incompatible? Had someone already accepted an offer? Each answer would suggest a different next step. A generic refusal leaves me guessing.

Without useful feedback, every revision starts to feel like negotiating with a locked door. I can keep changing the shape of the key without learning anything about the lock.

Hiring is competitive. Another person can be a better fit. Some replies explicitly say the position was filled by someone else. Those are understandable outcomes. What I find dim is a process that can ask for detailed accounts of my work and return so little detail about its own decision.

There are exceptions worth preserving. One recruiter responded specifically to my visual work, suggested a different role and invited me to explore it. That was useful. It gave me a direction instead of another reason to rearrange my résumé. Some employers also explicitly say humans reviewed the application; one states that it does not use AI for that review. The inbox cannot tell me that every refusal came from a machine, or that nobody looked at the work.

The broader market adds context. The Bureau of Labor Statistics reported a preliminary payroll increase of 29,000 in September and unemployment of 4.2 percent, describing both as little changed. Those national figures do not explain my individual outcomes, but they make it harder to treat every difficult search as a personal failure of effort.

Sources: S2

Recruiting teams face volume, too. Ashby’s April 2026 analysis found that applications per hire remained above 300 throughout 2025 in its dataset. That is a vendor cohort, not a measure of every employer or my odds. It does help explain why managing applications has become a substantial task in itself.

Sources: S3

I helped create that volume. I am one guy pursuing hundreds of roles. Some may have been a stretch; applying broadly does not make me the right candidate for every opening. I cannot celebrate the speed of my own AI workflow and pretend that the people receiving its output have unlimited time.

Hiring managers are probably right to use AI to organize and sort applications as it becomes mainstream. The responsibility is to make the criteria sensible, inspect mistakes and allow meaningful human judgment. Human in the loop is essential for the time being. The human needs enough authority and attention to notice when the filter gets it wrong.

The delusion is believing that a process becomes more intelligent simply because it becomes faster. An applicant can generate more polished explanations. A hiring team can process more polished explanations. We can both improve our productivity while learning very little about whether we should work together.

My next experiment is to put more thought into fewer, better-matched applications. AI makes it easier to build a short presentation around a particular posting: relevant work, the problem I think the role needs to solve, a proposed approach and what I would measure. Completed projects and proposed ideas need to be clearly distinguished.

A custom deck may make my thinking easier to inspect. It may also become one more attachment someone never opens. We have not demonstrated that it improves interviews or offers. I want to measure responses and conversations, instead of treating another batch of generated slides as a victory.

After all these submissions, that is what I want from the job hunt: enough information to make the next attempt more intelligent, and a real conversation when the work deserves one.

I have made myself easier to submit. I would like the hiring process to become better at understanding what has been submitted.

Prepared with AI assistance from application records and employer correspondence. Private correspondence, contact information and message identifiers remain unpublished.

Sources: S1 · S2 · S3

Why it matters

Faster application workflows produce more submissions without necessarily producing useful feedback or better hiring decisions. John N. Farmer argues for clearer evidence, meaningful human review and actual conversations.

Sources

1. I used AI to apply for hundreds of jobs. The next step is a better argument. — OMIKINA / John N. Farmer · 2026-10-10
2. The Employment Situation — September 2026 — U.S. Bureau of Labor Statistics · 2026-10-02
3. Recruiter Productivity Trends Report — April 28, 2026 — Ashby