Make who breeds part of the rule, so the divergence is attributable

The two runs allocated contested breeding slots in different orders: the
individual run by member insertion order, the aggregate run by stored class order,
which is youngest first. When eligible parents outnumber free slots that decides
which parents reproduce, for a reason that has nothing to do with averaging. The
control never reached the capacity boundary and so could not see it, which left
ADR-0040's central claim unsupported rather than merely imprecise.

The allocation order is part of the stated rule now, oldest first, honoured at both
resolutions: classes partition ages, so the oldest class holds the oldest members,
and ordering inside a class stays something only the individual run can see, which
is a resolution effect rather than an implementation difference.

A second control sits at the capacity boundary, twelve eligible parents contending
for one free slot, and both controls compare the whole distribution of members
rather than censuses. That change matters by itself: whoever breeds hands the same
investment to a child either way, so every total agrees while the distribution does
not, and a census-only control cannot see a different parent having bred. The
probe diverging only by tick 44 is that effect.

Attribution was verified by mutation rather than argued. Restoring the individual
run's insertion order fails the new control; restoring the aggregate run's
youngest-first order fails it; the earlier one-unit drift in aggregate upkeep still
fails the first. Since either allocation order alone fails, the reported divergence
is attributable to averaging.

My first attempt at the control was wrong and the same sweep caught it. I gave the
contested fixture ages descending with the index, reasoning that this made
insertion order differ from age order; it does the opposite, and the
individual-side mutation passed. Ages ascend now, so oldest-first disagrees with
both the insertion order and the stored class order. Without the sweep the new
control would have looked like it established attribution while covering half of
it.

Also added the depletion check the ADR's attribution argument relies on: a shared
store that ran out would make intake order-dependent too. I had written that the
fixture asserts this before it did, which is now true.

The measurements were regenerated and are unchanged. Saying so plainly rather than
as confirmation: the finding was about what the numbers can be attributed to, not
about their values.

Release 38/38, analysis, ASan 38/38 and clang-format all ran and passed here.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This commit is contained in:
dingo
2026-09-25 07:21:55 -04:00
co-authored by Claude Opus 5
parent b4c3da56e9
commit 55a728acd4
4 changed files with 253 additions and 24 deletions
+51
View File
@@ -1322,3 +1322,54 @@ The proof golden `d61ad5bf0c928c76`, the ecosystem fingerprint `2cd4a15f9fbe9949
the ensemble fingerprint `5ec59d390909b526` and the frozen genome fixture are
unchanged. No deployment manifest or image pin changed. Linux validation is pending;
no Gitea result is claimed since `b01a824`.
Review of `b4c3da5` found one defect (2026-09-25), now fixed with the controls
extended to catch it.
The two runs allocated contested breeding slots in different orders. The individual
run iterated members in insertion order; the aggregate run walked classes as stored,
which is youngest first. When eligible parents outnumber free slots that decides
which parents reproduce, and it does so for a reason that has nothing to do with
averaging, so the published divergence was not attributable to resolution. The
control never reached the capacity boundary and so could not have seen it, which is
the part that made ADR-0040's central claim unsupported rather than merely
imprecise.
The allocation order is now part of the stated rule, oldest first, honoured at both
resolutions: classes partition ages, so the oldest class holds the oldest members,
and ordering inside a class remains something only the individual run can see, which
is a resolution effect rather than an implementation difference.
A second control was added at the capacity boundary, where twelve eligible parents
contend for one free slot. Both controls now compare the whole distribution of
members rather than censuses. That change matters on its own: whoever breeds hands
the same investment to a child either way, so every total agrees while the
distribution does not, and a census-only control cannot see a different parent
having bred. The reported probe diverging only by tick 44 is exactly that effect.
The attribution was then verified by mutation rather than argued. Restoring the
individual run's insertion order fails the new control; restoring the aggregate
run's youngest-first order fails it; and the earlier one-unit drift in the aggregate
upkeep still fails the first control. Since either allocation order alone fails,
the divergence reported in the results file is attributable to averaging.
One of my own attempts at the control was wrong and was caught by the same mutation
sweep. The contested fixture first gave ages that descended with the index, on the
reasoning that this made insertion order differ from age order; it does the
opposite, since descending ages make insertion order identical to oldest-first, and
the individual-side mutation passed. Ages now ascend with the index, so oldest-first
disagrees with both the insertion order and the stored class order, and each
mutation alone is caught. Had the sweep not been run, the new control would have
looked like it established attribution while covering only half of it.
The measurements were regenerated and are unchanged: the same five widths, the same
two regimes, the same gaps. That is worth stating plainly rather than presenting as
confirmation, because the finding was about what the numbers could be attributed to
and not about their values.
Coverage limit added. Attribution rests on the controls covering every stage where
the two runs could order things differently. Breeding allocation is the only
order-sensitive stage, and the environment is sized so intake never depletes it,
which the fixture asserts rather than assumes. A future stage introducing another
order-sensitive decision would need its own control before these numbers could be
trusted again.
@@ -1,4 +1,4 @@
{"experiment":"resolution-equivalence-v1","model":"fixture-v1","compiler":"MSVC 1944","build":"release","commit":"m4.5","founders":100,"ticks":10000,"scarce_axis":"energy","lifespan":400,"capacity":200,"note":"Aggregate and individual runs apply one rule at two resolutions. The width-1 control matches exactly; these are the coarse measurements. Not an equivalence claim. See docs/adr/0040-resolution-equivalence.md"}
{"experiment":"resolution-equivalence-v1","model":"fixture-v1","compiler":"MSVC 1944","build":"release","commit":"m4.5-r2","founders":100,"ticks":10000,"scarce_axis":"energy","lifespan":400,"capacity":200,"note":"Aggregate and individual runs apply one rule at two resolutions, including the breeding allocation order, which is part of the rule. Two width-1 controls match exactly, one in ordinary growth and one at the capacity boundary, comparing whole member distributions rather than totals. Regenerated after the allocation order was matched; the values were unchanged. Not an equivalence claim. See docs/adr/0040-resolution-equivalence.md"}
{"class_width":10,"individual_living":75,"aggregate_living":100,"individual_energy":6275,"aggregate_energy":6120,"living_gap_per_thousand":250,"energy_gap_per_thousand":24,"aggregate_survived":true}
{"class_width":25,"individual_living":75,"aggregate_living":0,"individual_energy":6275,"aggregate_energy":0,"living_gap_per_thousand":1000,"energy_gap_per_thousand":1000,"aggregate_survived":false}
{"class_width":50,"individual_living":75,"aggregate_living":120,"individual_energy":6275,"aggregate_energy":6000,"living_gap_per_thousand":375,"energy_gap_per_thousand":43,"aggregate_survived":true}