A record you keep for yourself is allowed to be unflattering. Nobody is grading it. A bad week is information, an empty stretch is a fact about your month rather than an accusation, and the most useful entry is often the one you would not volunteer. Turn the same record outward and everything about it changes. Not the numbers. The numbers are identical. What changes is that the reader no longer has the context, and the context was doing most of the work. This is the seam between the two halves of the idea, and it is worth being precise about, because almost every way of crossing it badly looks reasonable from one side. The private half runs on a shared secret When you are the only reader, you and the record already agree about what happened. A gap in the calendar is a week you remember. A session that ran three hours and produced nothing that survived is one you can place. The evidence is thin on its own and it does not have to be thick, because you supply the rest. That is why honest-empty is cheap in private. A reading that says "not yet, there is not enough here" costs the reader nothing, because the reader knows why there is not enough. It is also why the unflattering entry is the valuable one. You are not defending it to anyone. The public half loses the reader who knew A stranger reading the same evidence has none of that, and will not leave the gaps empty. People do not read absence as absence. They read it as a claim. An empty stretch becomes a lapse. A partial month becomes a slow month. A number without its window becomes a permanent property of the person. None of those readings are in the data, and all of them are what the data will be used for if it is published as though it were self-explanatory. This produces the rule that most of the difficulty hangs off: whatever a value needs in order to be read honestly has to travel with it. The measurement window, what was covered, what was not, and whether a stretch is genuinely quiet or simply unobserved. A public number stripped of those is not a smaller truth. It is a different claim. Publishing evidence is not showing someone your record. It is making a much smaller and more careful statement, to a reader who cannot ask you a follow-up question. Which is why it is a projection, not an export The tempting design is a switch. One control, private or public, and the record goes out as it stands. It is the wrong shape, for a reason that has nothing to do with caution. An export answers a question nobody asked. Nobody wants your whole record, and you do not want to defend it. A projection answers the real question, which is narrower: which parts of this represent me. So the design Seorak holds to is selection all the way down. Nothing is public unless it is chosen. Profile fields, projects, which measured values accompany a project, and the measurement window are separate decisions, and where the projection can be reached, on the web, in search, over the API, over MCP, is a separate decision again. Publishing produces a frozen version rather than a live window, so the private record can keep changing without the public one moving underneath the person who published it. Withdrawing removes the projection and leaves the record it came from untouched. There is one thing the design does not do, and it is the part worth naming. A user-authored description of a project sits beside measured evidence, and Seorak does not check the description. It cannot. A sentence about why a decision was made is not the sort of thing a session record can confirm or deny, and a product that implied otherwise, by rendering the claim and the measurement as one thing, would be lending credibility it does not have. So the two stay visibly separate: this is what the person says, and this is what was measured. The objection Selective publication has an obvious problem and it deserves stating in full rather than in a closing caveat. The selection is invisible. A developer publishes their good months and their finished projects. A reader sees a projection and has no way to know what sits outside the frame, which means a curated record can mislead more efficiently than no record at all, precisely because it is made of true things. The honest answer is not that Seorak solves this. It does not, and no design does, because the problem is structural: any account of yourself that you control is an account you curate. What can be done is smaller. Every measured value carries its window and its coverage, so the frame is at least declared even when its contents are chosen. Absent evidence stays absent rather than being rendered as a zero, so a quiet period cannot be quietly converted into a measured one. There is no composite score, so there is no single number to arrange the selection around. That last one matters more than it sounds. Curation is most damaging when there is an obvious thing to curate toward. A record with no headline figure is a worse instrument for impression management, which is a strange thing to design for on purpose and the correct thing to design for here. None of that makes a projection complete. It makes it honest about being partial, which is the most a selective record can be. What this is actually for The reason to connect the two halves at all is not that a public record proves anything. It does not, and the last two sections were mostly about why. It is that the alternative is worse. Developers are already described publicly, by tool inventories, by job titles, and by activity counts that were never designed to say what they are read as saying. Against that, a projection that is chosen deliberately, states its window, and refuses to add up to a verdict is a modest improvement rather than a triumph. The private record exists so you can understand your own work. The public one exists so you can show some of it. Keeping those two sentences separate is most of the design, and collapsing them is the failure mode that every version of this idea has to be built against.