Registered before computing: collaboration overlap across an organization's projects¶

From the raw change records, which carry salted pseudonyms for each change's owner, submitter and attention set: for each C++ project in the committed source list, the set of accounts that appear on its changes, and for each organization the mean Jaccard overlap between the account sets of its different projects. No account is identified; only set sizes and intersections are read.

Prediction: AOSP's three projects (art, frameworks/av, system/core) share more of their people than Qt's six do, pairwise on average. If Qt's projects overlap as much as AOSP's or more, then collaboration density is not what separates a coherent organization from a loose one here, and the Tamura and Tsugawa mechanism does not transfer to this setting.

Outcome: refuted, and in the opposite direction.

Qt's projects share nearly twice as much of their people as AOSP's, and Qt's least-overlapping pair (0.140) matches AOSP's most-overlapping one (0.142). Yet AOSP is the coherent organization in update space and Qt is not. By the registered criterion, collaboration density is not what separates them here, and Tamura and Tsugawa's mechanism does not transfer to this setting -- which is no contradiction of their result, since they measure naming-style similarity across repositories by owner proximity, and this measures adapter-update similarity across projects by shared people.

The inversion is itself informative. The same engineers moving between qtbase, qtdeclarative and qt-creator do not make those projects' updates alike, while AOSP's three projects, staffed by largely different people, produce alike updates. What they share is a codebase: Android platform C++ under one style configuration, one lint regime and one build system. That points the same way as the probe did earlier today, where the organization was legible in code convention shapes and not in the reviewers' words: the fingerprint is a property of the codebase's conventions and tooling, not of the people who write to it. An organization is detectable when it imposes one set of conventions across its projects, whoever does the work.

That is three mechanism hypotheses for Qt refuted in a row -- pairwise coherence, centroid coherence, collaboration overlap -- and it is where generating a fourth would start to be fitting noise. What stands measured: AOSP is coherent at both lengths; Qt is not detectably coherent at either; longer local training makes Qt's grouping detectable anyway, and that survives matching composition. The mechanism for the last of those stays open, and the shared-conventions reading above is a hypothesis to test with the convention-shape probe on C++, not a finding.

(Account sets are taken from each change's owner, submitter and attention set, so bots that sit in attention sets count as accounts; small projects such as qtgraphs, with 62 accounts, make individual Jaccard values noisy. Neither changes the direction, which holds for every AOSP pair against Qt's mean.)