When both sides use AI, the CV is dead

The CV survived for one reason: it was expensive enough to be a weak signal of effort. Writing one took an afternoon. That afternoon is now thirty seconds, and the moment production became free, screening automated in response. The result is an arms race in which an applicant's model writes keyword-optimised prose for an employer's model to score.

Neither model knows whether the person can do the work. Both are fluent about it.

Fluency is not evidence

Large language models are competence engines for text. Feed one a job description and a thin work history and it returns a confident, well-structured account of a person who may not exist. This is not a misuse of the technology. It is the technology, doing exactly what it is for. The mistake is a hiring process that still treats polished text as a proxy for anything.

The legal system is already catching up with the other side of the race: automated screening is attracting discrimination claims, because a model trained on yesterday's hiring decisions inherits yesterday's biases at scale. A company that cannot explain why its system rejected someone is carrying a risk it has not priced.

The signal that survives

When text is free, what is scarce? Work that verifiably happened. A shipped project with a repository behind it. An automation a business actually ran. An outcome someone else can check. Evidence is expensive to fake in the exact way prose used to be, which makes it the natural successor.

Our position, plainly: we use AI here too, and we say where. It drafts write-ups from real repositories, and a human edits, owns and tags every word before anything publishes. AI drafts. Humans claim. Evidence scores. A capability on this platform cannot be typed into existence by anyone's model, including ours.