What changes and what does not

A CV was informative when writing one took effort. A cover letter was informative when it demonstrated that someone had read the posting. Both are now producible in seconds, so both carry less information than they did.

This is not new. Employers stopped weighting keyword-optimised CVs once optimisation became standard, and the same adjustment is happening to generated applications now.

It is a general property of signalling rather than anything specific to hiring. A signal works because it is costly to produce for people who lack the underlying quality, and anything that makes it cheap for everyone destroys it — which is why the response is always to find a new costly signal rather than to improve the old one.

Signals that remain costly to produce: work someone actually did and can be examined, a person who will vouch for them, a conversation that cannot be scripted in advance, a public track record built over years.

None of these is new either. They are what hiring relied on before documents became the default screening mechanism, and they are what it returns to as documents become unreliable.

That explains why hiring processes have grown longer rather than shorter while screening got faster. The cheap stage is now nearly free and proves nothing, so the expensive stages expand to carry the weight the document used to.

• Demonstrable work — code, writing, projects, shipped things
• Referrals from someone who stakes their credibility
• Real conversations, unscripted
• A public record accumulated over time

If applying costs nothing, employers receive far more and read each one less. That pushes them toward mechanisms that limit volume naturally — referrals, invitations, smaller channels, application fees in some markets.

It is an unappealing dynamic for candidates without networks, and it is the predictable consequence of making applications free to produce.

Some employers are going the other way, and it is worth noticing why. A very short application raises volume further and makes a small work sample the real filter, which is cheaper to evaluate at scale than a long document nobody trusts and produces a signal candidate automation cannot fake.

The shift towards referrals and networks is not neutral. Those are precisely the resources that are unevenly distributed, and a market weighting them more heavily advantages people who already have access.

Career changers, people without industry contacts, returners and anyone geographically distant lose the most when the open application stops working. The cold application was an imperfect channel and it was the one available to everyone.

Public work is the substitute that is actually buildable without a network. It is verifiable by a stranger, it does not require knowing anyone first, and it is the closest available replacement for a referral — which makes it the most rational place for an outsider to spend effort.

Candidates automate applying, employers automate screening, and the net effect on any individual’s chances is close to neutral while the total effort in the system rises.

The candidates who do well are not the ones who automate hardest. They are the ones who use automation to find better-fitting opportunities and then invest real attention in a smaller number of them.

It is a collective action problem rather than a mistake anyone is making. Everyone would be better off if applications were fewer and more considered, and no individual benefits from unilaterally sending fewer — which is exactly why it does not resolve on its own.

Build things that can be pointed at. Maintain relationships before you need them. Apply to fewer roles with more care. Use automation for discovery and administration, where it is unambiguously good.

This is unexciting advice and it is what follows from the dynamic. As the automatable parts of an application stop carrying information, what remains is what always did.

One cheap move outperforms most others: a short message to someone who actually works on the team, sent alongside the application. It costs ten minutes, it is not automatable at volume, and it converts at a rate that makes the arithmetic on mass applications look absurd.

And hold predictions lightly, including these. Deployment is slower than capability, the confident forecasts made when the first screening systems arrived mostly did not happen, and the strategies above pay off either way — which is what makes them worth choosing.