The rules files against no file and against each other, at penalty 1.0¶
# research(2026-10). rules-reading-openstack-rp1.0.json, rules-reading-wikimedia-rp1.0.json. The
read contrast, distilled own minus sibling, keeps its reading at penalty 1.0 (the rescoring entry
above). The contrasts the 2026-10-07 read reported but did not read move, and three now lie wholly
beyond the SESOI. Each is a 95% interval over the first ten admissible partitions:
| contrast | OpenStack, 1.1 | OpenStack, 1.0 | Wikimedia, 1.1 | Wikimedia, 1.0 |
|---|---|---|---|---|
| distilled own minus none | -0.0105 [-0.0215, +0.0006] | -0.0241 [-0.0401, -0.0089] | -0.0013 [-0.0074, +0.0043] | -0.0107 [-0.0207, -0.0014] |
| written own minus none | -0.0112 [-0.0230, +0.0013] | -0.0240 [-0.0396, -0.0096] | +0.0011 [-0.0051, +0.0073] | +0.0197 [+0.0059, +0.0345] |
| written own minus written foreign | -0.0039 [-0.0147, +0.0065] | -0.0274 [-0.0455, -0.0120] | +0.0056 [+0.0012, +0.0116] | +0.0397 [+0.0247, +0.0564] |
| distilled own minus written own | +0.0008 [-0.0075, +0.0082] | -0.0001 [-0.0080, +0.0084] | -0.0024 [-0.0070, +0.0012] | -0.0303 [-0.0452, -0.0167] |
At the registered decoder a distilled file lowers exact match on both organizations, both intervals below zero. The written guides split: Wikimedia's raises exact match over no file and beats OpenStack's guide on Wikimedia code by about four times the SESOI, while OpenStack's lowers it and loses to Wikimedia's guide on OpenStack code. These remain reported, exploratory contrasts: the rule fixed before the run reads only distilled own minus sibling. That a written guide can carry an organization-level difference the distilled files do not is the question the comparator was added to ask, and on Wikimedia it now has a development-window answer to test.