Applying got free. Reading didn’t.

Effort used to mean something

An application used to cost the person sending it an evening. That cost was the point. A cover letter written for one company told the reader that someone had read the posting, which is close to the only thing a stranger’s application can prove.

Language models made the evening free. Anaïs Galdin and Jesse Silbert studied a large freelancing marketplace before and after it added an AI writing tool. Afterwards, a tailored proposal stopped predicting who would do the job well. The effort was still on the page. It no longer told the employer anything.

So candidates send more, because each one costs nothing. Employers filter harder, with software, because there are too many to read. When The New York Times asked Gen Z readers about their job searches in October, the complaint was not only that hiring is slow. It was that they could not get “a real, live human being” to speak to them.

“Both massive, speculative bulk applications and aggressive automated filtering are now very easy. And neither employees nor applicants seem to be happy with the new equilibrium.”

Friction alone won’t fix it

One reply to that thread suggested charging a cent per application. The instinct is right: make each application cost something, so people send the ones they mean. But 511 applications at a cent each is $5.11. That stops nobody who already pays for an auto-apply tool, and it does stop a broke 22-year-old.

The other easy answers fail the same way. A required essay box costs a bot two seconds and an honest candidate ten minutes. Blocking paste breaks dictation and screen readers, and bots type through it anyway. AI detectors are wrong often enough to be useless, and they flag writing by non-native English speakers far more often. Each one charges the sincere applicant more than the automated one.

The cost has to be one a bot cannot pay: something only the candidate knows, said somewhere they will be asked about it.

What we are building

One question per posting, about the work the role actually does, answered before the application goes in. For an operations analyst role whose posting says you will chase down the outliers in a weekly carrier scorecard:

Tell us about a time a number in a report looked wrong to you. What did you check first, and what did you find?

A model can write an answer to that. It cannot sit the conversation afterwards. The candidate is told up front that the first interview starts from their answer, so a made-up example gets past the form and not much further.

It works where hiring already happens. On a form you own, the question sits inside the form. On a form your ATS hosts, the candidate answers on a short hosted page and gets a response code, which goes into a required question on the application. Each code is tied to one posting and one email, and you can check it.

It comes with a deal. Your team commits to opening every answer within a set window, and candidates see your actual open rate, counted from real opens rather than typed in as a promise. They also keep a copy of what they wrote, can withdraw it, and hear from us once when someone on your team opens it.

What we will not build

• Scores, rankings, or summaries of answers. People on your team read them.
• AI detection. Candidates say whether they used AI, and the honest answer is the signal.
• Sending answers to an AI model, or using them to train one.
• Paste blocking or keystroke tracking.

Software that scores the answers would make us the faceless screener this whole problem is about.

Design partners: what we are asking for

None of this is live yet. We would rather build it with the teams who will use it than guess, so we are looking for five.

You are a fit if

• You hire for entry-level or early-career roles.

  That is where the volume is worst and where 10xCareer already works with candidates.

• Your postings get more applications than your team can read.

  If you read every application today, you do not need this yet.

• You will open every answer within 14 days on the pilot postings.

  This is the one that matters. Candidates see whether you did.

• You will tell us what works, on a short call every two weeks.

  Which ATS you use, which channels bring the volume, and where the console gets in the way.

What you get

• Free use during the pilot.

  No contract and no card. Pricing comes after, and partners hear it first.

• Your ATS first in line for a direct integration.

  Partners set the order in which we connect answers to the candidate record.

• A say in the defaults every later team gets.

  The question editor, the open window, the notice candidates read, and the console.

One condition, stated plainly: if fewer than three teams will commit to opening the answers, we will not build this. A question nobody reads is one more wall, and candidates have enough of those.

If you are job hunting

Nothing changes for you today. When this ships, the question will be short, it will say who reads it and by when, and you will keep a copy of your answer. Until then, scan a posting before you apply, so the applications you send are the ones you mean.