Evidence beats résumés

The résumé was always a proxy

A résumé has never been the thing an employer actually wants to know. It is a proxy for it, a compressed, self-reported claim that a person can do the work. For decades the proxy held, not because it was accurate, but because it was costly. Writing a coherent account of your own work took effort, and the effort itself carried information: this person can organize their history, judge relevance, and communicate under constraint.

Proxies survive exactly as long as they stay expensive to fake. The moment a signal can be manufactured at no cost, the market that priced it collapses. That is not cynicism; it is how every signal in every market has always worked.

Polish just became free

That is what changed. A fluent, tailored, keyword-tuned document can now be produced in seconds by anyone, from any starting point, for any purpose. The floor of polish has risen to the ceiling. And when every document in the pile reads well, reading well tells the reader nothing at all.

This is not only a hiring story, though hiring is where it is most visible. The same collapse is happening to the weekly status update, the project proposal, the vendor questionnaire, the performance self-review, the incident write-up. Every artifact whose credibility rested on being well written is now cheap to produce and therefore weak as evidence. Readers adjust: they trust the polished document less each quarter, and start asking for the thing behind it.

None of this means people got worse or that the tools are bad. It means a proxy died the way proxies always die: by becoming too easy.

“When polish costs nothing, polish means nothing.”

What still costs something

So the question is not how to write a better document. The question is: what remains expensive to fake? The answers are unglamorous and always have been. Showing up repeatedly, over weeks, is expensive to fake. Work performed under observation is expensive to fake. Artifacts with a date and a history, this was built, here, on these days, and it ran, are expensive to fake. A capability demonstrated on demand is expensive to fake.

Notice what these have in common: they are records of behavior, not descriptions of it. A description can be generated. A dated trail of work, decisions made, problems attempted, difficulty climbing over time, artifacts accumulating, can only be produced one way, which is by doing the work. The cost is the credibility.

This is the signal that survives, because it is the only kind left standing. Not credentials, which certify a past exam. Not endorsements, which are social courtesy at scale. Evidence: specific, dated, verifiable, and small enough to check.

A claim is a claim. Evidence is a receipt.

The distinction is worth making precise. A claim says: I am proficient in this, or the migration went well, or the assistant is handling those tickets. A receipt says: on these dates, here is what happened, at this level, and here is the artifact. Claims invite skepticism and require interrogation. Receipts invite verification, which is cheaper for everyone.

For the person doing the work, this reframing is quietly liberating. You cannot control whether a claim is believed. You can absolutely control whether a receipt exists, because receipts are manufactured by practice, and practice is available to you today, in the job you already have. Every real piece of work you finish and can explain is a fact about you that no one can inflate away.

It also changes what you are competing on when the machine can write anything you can write. If the document is free, the document is not the differentiator. What you checked, what you refused, and what you can still explain a month later is the differentiator, and none of those are produced by a better prompt.

Stop polishing. Start proving.

The practical consequence fits in two sentences. Every hour spent re-sanding language that a machine can generate is an hour spent competing in a market where the price of the good is falling to zero. Every hour spent doing and recording the work, building, deciding, checking, shipping something small and dated, mints an asset whose value is rising, because scarcity is moving from polish to proof.

This is also the saner way to work. Polishing is anxious work with no floor; the document is never done because it was never the real problem. Practice has edges. A session ends. A finished call counts whether or not anyone replies to the email that week. Progress becomes something you did rather than something you await, and that difference is worth more than morale: it is the difference between a year that erodes you and one that builds inventory.

“A résumé is a claim. Evidence is a receipt.”

Where Datum stands

We should say plainly where we sit in this argument. Datum's practice emits proof: small, dated, verifiable records of what was actually decided and demonstrated, accumulated as a side effect of doing the work, stored in a vault Datum cannot read, and shared only when the person holding them consents. It is one implementation of the idea, and the idea matters more than our version of it. If the industry converges on proof over polish through other doors, the person doing the work still wins.

The résumé had a long and honorable run as a proxy, and the same fate is arriving for every other document that was trusted because it read well. The conditions that made them informative are gone, and they are not coming back. What replaces them will be built from receipts. Start collecting yours.