I had been treating the distinction between an idea and a project as though there ought to be a threshold: enough notes, enough implementation, enough commitment, enough certainty. That is administratively convenient and descriptively poor.
The more convincing mechanism is recurrence. A minor observation catches attention. Later, an unrelated problem gives it another edge. A discarded design suggests a use somewhere else. A company name resembles an old mechanism. A single word in a document becomes worth searching because it appears to touch several things that were previously separate.
None of those moments establishes importance. Most small curiosities will remain small. The useful signal is that some of them keep producing further questions without being forced to do so.
An idea becomes a project when it keeps giving you reasons to come back.
This is different from choosing a project. Work can begin by decision: build this feature, solve this fault, document this system. But another class of project is recognised retrospectively. By the time it has a vocabulary, contradictions, dependencies, abandoned branches and unresolved questions, the declaration is almost redundant. The structure already exists.
That also means relevance cannot be fully assigned in advance. A rejected mechanism, a stray name, an obsolete drawing or an unexplained room may have no present value beyond being interesting. Later information can alter that relation. What looked peripheral may become the missing hinge between two larger bodies of work.
The correct behaviour for the Station is therefore not to preserve everything because everything is secretly important. It is to avoid demanding proof of importance before curiosity is allowed to continue. The cost of another note or another room is sometimes lower than the cost of prematurely deciding what the future will not need.
This is also why abandoned laboratories should remain physically old rather than being rewritten into archives. A current catalogue explains the past from the present. An old bench preserves the assumptions that existed before anyone knew which parts would survive. Returning to it can expose possibilities that a cleaned history would have quietly removed.
There is no algorithm here that decides what deserves pursuit. Interest is not evidence. Recurrence is not truth. But repeated voluntary attention is information about the investigator as well as the subject. It reveals where the unresolved structure is exerting pull.
The practical rule is modest: follow the piece that is interesting enough to inspect again. Let the pattern appear after the fact.
Filed from the Station after discussion of CEU descendants, dormant technology, old laboratories and the point at which an idea ceases to be merely an idea.