App to Stack / Field Note 22
The Capability Receipt App to Stack
The next credential is not proof that a person used AI. It is a readable account of the judgment they brought to the work.
01 / The missing proof
A finished answer is an inadequate credential.
As AI moves from a helpful application to a set of capabilities people assemble around their own work, the old evidence of competence gets weaker. A portfolio can show an attractive outcome. A certificate can show attendance. A prompt can show a moment of instruction. None of them reliably shows what a person noticed, what they rejected, what constraint they protected, or whether they could improve the result when the next situation changes.
A capability receipt is a compact, inspectable record of the judgment, constraints, revisions, and results behind an AI-assisted piece of work. It is not a surveillance log. It is not an invitation to hand an employer the private memory, personal sources, or reusable methods that make up someone’s stack. It is evidence designed to travel without turning the person’s entire capability into a take-home assignment.
That distinction matters because the app-to-stack transition changes what work looks like from the outside. An App Person is often legible through the application they were trained to operate: the tool names the task, the company owns the workflow, and the credential names the menu. A Stack Person may combine research, retrieval, drafting, automation, judgment, and review in a method that does not appear inside any one product. The method is real even when the tool list is not enough to describe it.
02 / Three kinds of evidence
The artifact, the trace, and the claim are different things.
Hiring and education tend to bounce between two bad substitutes. One is the polished output: here is a strategy memo, a campaign, a model, a lesson plan. The other is exhaustive process capture: show every prompt, tab, draft, and token. The first hides too much. The second asks for too much. A capability receipt creates a smaller third object.
The receipt is a claim that can be questioned. “I chose this audience because the customer interviews kept naming this failure mode.” “I excluded these examples because they were confidential or outside the stated scope.” “The first draft optimized for speed; the revised version added a human approval step because the action could not be safely reversed.” Each statement gives a reviewer somewhere intelligent to look.
03 / What travels
Portable proof should be thinner than private capability.
The temptation will be to make every receipt comprehensive, standardized, and easy to score. That would recreate the application era’s familiar error: taking a living practice and flattening it until an institution can process it. A receipt should be just specific enough for the decision at hand. A school may need to see how a student checked an answer. A prospective client may need to see how a consultant bounded a recommendation. A manager may need to see what approval was retained before an agent acted.
Those are different questions. They deserve different receipts. The consistent element is not a rigid form but a social promise: the person can make their reasoning inspectable without pretending that their whole stack is public property. This is especially important for people whose capability has been compounded over years. They should be able to show growth without exporting the personal context that made the growth possible.
- A clear claim about what the person was responsible for.
- The relevant constraint or boundary, stated without disclosing protected context.
- One or two consequential revisions that reveal judgment rather than busywork.
- A result another person can inspect, challenge, or test.
04 / A new bargaining surface
Institutions will have to learn how to read these.
For employers, a capability receipt could make interviews less theatrical. Instead of asking someone to perform intelligence in a blank room, a team can discuss a real decision: what was the alternative, what made it unsafe, what changed your mind, and what would you do with a different constraint? That conversation tests judgment more directly than whether someone can produce a clean first answer under artificial conditions.
For educators, it offers a way out of the false choice between banning assistance and accepting unaccountable output. The question becomes whether the learner can name their contribution, explain a revision, and defend the standard they used. For independent workers, it may become a way to build trust without giving away the machinery that makes them valuable.
There is a risk, of course. Receipts can become a new compliance ritual, filled with boilerplate and graded for polish. That is why the useful unit is not the form. It is the question the form invites: can you show enough of your judgment for someone else to rely on it? Stack People need an answer that is portable, bounded, and alive to context.