7 min read
Let me say the uncomfortable thing first, because everything else follows from it.
Your degree is not evidence that you can do the work. It is a receipt. It records that you attended, that you sat the exams, and that an institution with a reputation to protect is willing to vouch for that narrow claim. It was never built to tell an employer whether you can run a pricing analysis, write a campaign brief, or clean a dataset without being told twice. For decades we behaved as though it did, because nothing better existed and because the pretence was cheap for everyone involved.
The pretence has stopped working. What broke it is not visa policy, not a soft graduate market, and not employers becoming unreasonable. What broke it is that the artefact became free.
The output is free now. The judgement is not.
A student can produce a competent market analysis in twenty minutes. A clean deck. A tidy financial model. Reasonable copy. Working code. Not brilliant, but presentable — and presentable is exactly the standard a portfolio used to be judged against.
So the portfolio died the moment the deliverable became free to produce. Every recruiter knows it. Most have stopped looking at links. They are not lazy; they are rational. An artefact that anyone can generate carries no information, and a signal that costs nothing to fake stops being a signal. That is not a complaint about AI. It is arithmetic.
What did not become free is the judgement. Somebody who has shipped this kind of work, who has hired for it, who knows what the client actually wanted and what the student misread — that person’s opinion still costs time, and still carries information. The scarce thing moved. It used to be the output. Now it is the assessment of the output.
Almost nobody has repriced around that. Certificates still get issued for watching videos. Simulations still get auto-marked by a rubric nobody signed. Badges still get awarded by platforms rather than people. All of it was mildly useful when producing work was hard. None of it survives contact with a world where producing work is trivial.
Four conditions, and a stamp without all four is decoration
A proof of work is trustworthy when someone’s name is attached to a judgement, and that person can be found, questioned, and embarrassed. Everything else is packaging. Concretely, four conditions:
A named assessor with something to lose. Not “assessed by our platform”. Not “reviewed by industry experts”. A person, named, who works in the market the student is trying to enter and who will still be working in it next year. A judgement is only as good as the reputation standing behind it, and an anonymous judgement has no reputation standing behind it at all. When a Nap OS brief comes back, it comes back with a name on it — a Project Manager with five or more years in that domain, working in the destination market. If their assessments turn out to be worthless, that is their name on the worthless assessments.
Criteria published before the work starts. This is the condition most people skip, and skipping it is how assessment becomes theatre. If the marking criteria appear after the submission, the assessor is describing their taste, not measuring performance. Publishing criteria in advance costs the assessor the freedom to move the goalposts, which is exactly the point. The student knows what they are being measured against before they start. So does the employer who reads it later.
A second marker, and an agreement rate you are willing to publish. If two qualified reviewers assess the same work against the same published criteria and reach different conclusions, the criteria are not doing any work and the stamp means nothing beyond one person’s mood on one afternoon. The only honest way to know is to measure it — sample the work, mark it twice, count how often the two agree, and publish the number. Most assessment providers do not do this. Some do not do it because it is expensive. Most do not do it because they are afraid of the answer.
A record that travels and can be inspected. The student must own it, carry it to any employer, and show the whole chain: the brief they were given, the deliverable they produced, the criteria it was measured against, and the name of the person who measured it. If an employer cannot inspect the work itself and can only see a badge, you have not replaced the CV. You have built a smaller, shinier CV.
Now the part that is ruthless about us
If those are the conditions, we should be held to them in public, including where we do not yet meet them.
We have not measured our reviewer agreement rate. Not once. We will measure it for the first time by 31 January 2027, and we will publish what it says — including if it is bad. A proof-of-work company that will not show its own error rate is asking for exactly the unearned trust it claims to be replacing.
We do not guarantee employment, interviews, or any immigration outcome, and we never will. We hold an Irish employment agency licence, EA 5651, which among other things means we may not charge a jobseeker for finding them work, and we do not. A stamp is not a job. Anyone selling you a stamp as a job is selling you something they cannot deliver.
We are not an accredited award and we do not want to be. Academic credit and academic supervision belong to your institution. We supply one input — industry assessment — and the institution decides what to do with it.
And the hypothesis underneath all of this is still a hypothesis. We believe a student who arrives holding work assessed in the destination market gets hired sooner than one who does not. We think the mechanism is obvious. We have not proven it at scale, and until we have, we will describe it as a belief rather than a result. Two cohorts, measured properly, and then we will either show the number or stop saying it.
That is the standard. It is not a modest one, and we are not currently at the top of it.
Why a Project Manager, and not a professor or an algorithm
Not because academics cannot assess — they assess better than almost anyone, against the thing they are assessing for, which is academic merit. That is a different question from the one an employer is asking.
The employer’s question is narrow and slightly brutal: if I gave this person this task on Monday, what would land on my desk on Friday, and how much of my own time would it cost me? The person best placed to answer is someone who has given that task to someone and lived with what came back. They know which shortcut will hold and which will collapse under a client question. They know that the deck was fine and the recommendation was cowardly. An algorithm marking against a rubric cannot see cowardice in a recommendation. A practitioner can, and will say so in writing, within seven days.
This is also why the assessor has to work in the market the student is moving to. Competence is not universal; it is local. How you push back on a manager, how much hedging reads as thoughtful rather than evasive, when to send the email and when to walk over — none of that transfers cleanly across borders, and all of it is being judged in the first ninety days of a job.
What this actually changes
For the student: you stop asking to be believed and start handing over something inspectable. That is a worse deal if your work is weak, which is the point. The record is yours, it is portable, and nobody can take it back.
For the institution: employability stops being a promise made at recruitment and becomes an outcome with a paper trail — contracted, reported, and reportable regardless of whether anyone gets hired. You are no longer vouching for something you cannot observe.
For the employer: you look at the work before you look at the CV. Not a proxy for the work. The work, the criteria, and the name of the person who judged it — and if you think that judgement was wrong, you can go and ask them why.
Trust has never scaled by being asserted. It scales by being attributable. A credential says an institution believes this. A proof of work says this person did this task, measured against these criteria, and this named practitioner put their name to that assessment.
One of those statements can be checked. That is the entire difference, and it is enough.
Nap OS is built by Napblog Limited, Dublin — CRO 812006, Licensed Employment Agency EA 5651. Five project domains: Startup & Entrepreneurship; Strategy & Operations; Marketing & Sales; Product & Research; Data & AI. os.napblog.com
