Applied AI · Evidence of Merit
Thesis stage
Replacing Resumes
The observer that already watched the work speaks for it
What this is
Resumes are now job-description keywords, and from the outside there is no way to tell which ones still represent anyone at all. Replacing Resumes proposes the exit: a portable record of real working interaction — distilled from the long collaboration between a person and the LLM they build with every day — abstracted into project-based evidence of merit: an authenticated reference from the observer that watched the work happen.
The technical shape
- The problem
- resumes collapsed into job-description keywords — representation anywhere in 1–99%, by our estimate
- The cycle
- volume → stricter keyword gates → misrepresentation → more gaming → stricter gates — unsolvable from inside the loop
- The witness
- the LLM one works with daily — no emotional stake, full contextual continuity
- The product
- a portable record of real working interaction, abstracted into project-based evidence of merit
- Stage
- thesis — all the pieces are present for implementation; the implementation is the project
The deep end · full technical outline
Outline v2 · expanding
What the résumé has become
Resumes are now job-description keywords. ATS sorting made them that way — and the tracking systems are not the villain: the internet lets a person apply to far more jobs than they ever could before, no AI required, and no human screen can absorb that volume. AI does not change the shape of the problem; it turns the dial, making the ATS more necessary for any business trying to keep up. Where a résumé may once have represented an individual to something like 90–99%, our estimate — and it is only an estimate — is that it now lands anywhere between 1% and 99%. The document and the person have come apart.
The doom cycle
There is not really any blame to be directed here — each factor is a reasonable response to the one before it. Rising application volume pushes employers into ever-stricter keyword-based acceptance; strict keyword gates are the root cause of misrepresenting the applicant; and misrepresentation invites more gaming, which invites stricter gates. It is a negative-reinforcement cycle, and we believe this state is unsolvable from inside the loop — it will keep driving the bell curve of misrepresentation toward the higher threshold. Resumes, in the traditional sense, are over.
Where hiring retreats
This doom cycle is the primary driver behind the rise in who-you-know hiring. A former colleague is at least evidence of something — but a channel built on connection still undercuts a meritocracy, and that degrading effect costs businesses in the long run, if not the short: a hire made on connection alone can arrive carrying no signal at all, and the business absorbs the mismatch before the arrangement finally ends. Hiring retreats there anyway for a simple reason — informational survival. When the synthetic signal has become indistinguishable from the authentic one, already knowing someone is the last channel that cannot be gamed from the outside.
Merit versus known
What the who-you-know hire actually reveals: the difference between a known candidate and an unknown one takes far under six months of working together to derive. In one case the candidate is, in some sense, already known to the business. In the other, there is a keyword document. So the question becomes: how do you make an authentic, accurate representation of a candidate known to a business — without the six months, and without the bias?
The witness
With whom does one work consistently, and over a period long enough, to render the output that actually matters — a merit-based, authentic representation? Family and friends are out, on bias — bias is precisely the thing that corrupts the who-you-know channel in the first place. An ATS is out for the opposite defect: no context at all, only keyword symmetry. Which leaves the collaborator many of us now work beside every day: the LLM. It has worked with us consistently and over a long period. Unlike family and friends, it has no emotional stake in the outcome. Unlike the tracking system, it has contextual continuity — it has seen the process, the revisions, the doubts, the dead ends, and the deliveries. So turn the perceived negative into the best possible solution: let the observer that already watched the work speak for it.
What this project builds
The shape of it: a portable record of real working interaction, abstracted into project-based evidence of merit — an authenticated reference from an observer that watched the work happen. Not the raw transcript; a distillation of it — what was built, how problems were approached, what was revised, what shipped — able to say the one thing no keyword field can: hire this person because I have observed them build, debug, question, refine, and deliver — on actual work, not claims. Through implementation, a business gains an unbiased, merit-based review of an applicant. The doom cycle loses its grip, because the channel it corrupted stops being the channel that matters.
The honest stage
This is a thesis, and it is presented as one. No working instrument is claimed here yet — what is claimed is the argument: the diagnosis of the résumé’s collapse, the reason the cycle cannot be solved from inside, and the identification of the one observer that is both unbiased and fully contextual. We believe all the pieces are present for implementation. The implementation is the project — and hiring returns to a representative format: the person whose ability could never survive compression into one page of keywords finally gets represented by the work itself.
Related in the lab
- CodeRadar →
- Shares instrumentation — building the instrument that makes a hidden state legible.
- IMAXING Latent Space →
- Shares instrumentation — building the instrument that makes a hidden state legible.
Built at this bench
The machine, the instrument, the film that explains it — all designed and built in-house. The same bench works for hire.
Services →