Skip to content
Sign in
← All notes

The end of the perfect CV, and what replaces it

When every application is well written, well structured and tailored to the role, the application stops carrying information. Hiring has to move to evidence, and most of it has not.

Hiring7 min read

A CV was never a description of somebody. It was a proxy, and the thing it proxied for was effort: whoever wrote a sharp, tailored, well-structured application had probably cared enough to do the work.

That proxy has collapsed. A tailored application now costs a minute, so a good one tells you almost nothing about the person who sent it.

What actually broke

Three signals recruiters relied on all fell at once. Writing quality, which now says nothing. Tailoring to the role, which is a paste operation. And the covering letter, which was always the weakest signal and is now the emptiest.

The volume changed too. Applying is cheap enough that applying widely is rational, so the pile got larger and less filtered at the same moment its contents got harder to distinguish.

The common response is to filter harder on keywords and years, which selects for whoever optimised best rather than whoever can do the job. That is not a fix, it is the same failure with more steps.

A pile of applications that are all well written, all tailored and all indistinguishable, next to the signals that used to separate them and no longer do.
The application used to be a proxy for effort. Effort is now a minute.

What still carries information

Everything that is expensive to fake, which now means everything that involves specifics, consequences, or a live conversation.

  • Things somebody built that you can look at. A repository, a portfolio, a product, a piece of published analysis. The artefact may be assisted; what it demonstrates is choices.
  • Decisions with outcomes attached. Not what they were responsible for, but what they decided, why, and what happened. Generated applications are strikingly bad at this because the model does not know what happened.
  • Specific detail about a real situation. Ask about a problem they were handed and follow up twice. Depth is the tell.
  • How they think out loud about a problem they have not seen. Nothing prepared survives the second follow-up question.
The evidence that is expensive to fake: something built that can be examined, a decision with its outcome attached, and a live conversation about an unfamiliar problem.
Specificity is the part that could not have been generated.

What this means if you are applying

The counter-intuitive part is that using AI to polish your application is now the low-value move, because everybody is doing it and it makes you less distinguishable rather than more.

What stands out is specificity that a model could not have invented: the actual decision, the actual number, the actual thing that went wrong and what you did about it. That is genuinely yours and cannot be generated on your behalf.

  • Lead with decisions and outcomes rather than responsibilities.
  • Be concrete about what changed as a result of you being there.
  • Attach something somebody can look at, even a small one.
  • Do not spend your effort on polish. Polish is free and worth what it costs.

Common questions

Are CVs still useful in the AI era?
As a factual record of where somebody has been, yes. As a signal of quality or effort, much less than before, because the writing quality and tailoring that used to indicate care now cost a minute. It is best treated as context rather than evidence.
Is it bad to use AI to write your CV?
It is neither wrong nor much help. Everybody is doing it, so it does not distinguish you, and it tends to smooth away the specific detail that would have. Use it for structure and grammar; supply the specifics yourself, because those are the part that carries information.
How should companies screen candidates now?
By things that are expensive to fake: work you can look at, decisions with outcomes attached, specific detail about real situations, and live conversation about an unfamiliar problem. Keyword and years filters select for whoever optimised best, which is a different quality from doing the job well.

PDP Quest exists because of the problem underneath all of these: when output stops indicating capability, you need another way to know who can actually do the work.

See how verification works →