Chromium yields reviewer comments at Qt's rate, and the two live organizations are stable to the seal¶

# research(2026-09). Two measurements, both taken because the AOSP collapse made them necessary.

The yield a candidate host actually delivers. Volume is not the quantity the corpus is built from: AOSP produced 5,133 usable examples while dropping 15,880 author comments and 7,305 unanchored hunks. What matters is inline comments that carry a line anchor and were written by somebody other than the change owner. Measured directly against the Gerrit API over 2025-10, nothing written into the corpus:

Chromium's review culture yields at least as well as Qt's, and chromium/src alone merges hundreds of changes a month, so a Chromium arm is not volume-limited. Two caveats stand. The design refuses an organization of one project, since it could not then be told apart from that project, so a Chromium arm means several of its projects rather than chromium/src alone. And adding a host is a corpus-scope decision that touches the ethics determination, which the report says is initiated through the institution rather than asserted by the plan.

The two organizations the study already has are not going the way AOSP went. Merged changes a month, from the fetch records, through the last month before the seal:

Neither trends down, and neither shows anything like AOSP's ninety-five percent drop. This is a check on the sealed window that does not spend it: the test window cannot be inspected, so the best available evidence that it will carry data is that change volume is steady right up to its boundary. It bounds volume only. Whether the content of those months resembles the training window is a separate question, and the drift probes are what answer it.