@@ -359,7 +359,7 @@ separates exact finite verification from retained all-size hypotheses.
 | [DaysSky](https://github.com/DaysSky) | [#150](https://github.com/CrocSwap/integer-mult-bounds/pull/150): delayed bit dirty reads, physical register recycling, exact scalar and frame checks. |
 | [William Porter / hpst3r](https://github.com/hpst3r) | [#147](https://github.com/CrocSwap/integer-mult-bounds/pull/147): terminal elimination on the shared bit word. |
 | [Thomas Marchand / Th0rgal](https://github.com/Th0rgal) | [#146](https://github.com/CrocSwap/integer-mult-bounds/pull/146): telescoping min-cut gauge subset, independently composed with the new schedule. |
-| [Rohan Gupta / gupt1156](https://github.com/gupt1156) | [#148](https://github.com/CrocSwap/integer-mult-bounds/pull/148): tighter atom exponent and adapter toll. |
+| [Rohan Gupta / gupt1156](https://github.com/gupt1156) | [#148](https://github.com/CrocSwap/integer-mult-bounds/pull/148): tighter atom exponent and adapter toll. The [source-retention receipt](research/pr148-source-retention/README.md) recovers 14 historical source gaps and separates the 11-file direct change from inherited work; whole-PR review remains separate. |
 | [James Chang / jamesyc](https://github.com/jamesyc) | [#124](https://github.com/CrocSwap/integer-mult-bounds/pull/124): compensated birth-read reuse principle used by #150. |
 | [SovereignSteak](https://github.com/SovereignSteak) | [#122](https://github.com/CrocSwap/integer-mult-bounds/pull/122): terminal accumulator elimination used by #147; [#151](https://github.com/CrocSwap/integer-mult-bounds/pull/151): independent matching gauge/recycling composition and the selected 473/10^6 atom exponent. |
 | [Andrew Barnes / Bortlesboat](https://github.com/Bortlesboat) | [#101](https://github.com/CrocSwap/integer-mult-bounds/pull/101): binds literal scatter operations to paid incidences and rejects malformed or unchecked inputs. |
 
@@ -0,0 +1,47 @@
+# Local task card (autoresearch.md section 3)
+
+Task ID / proposed scope key: `bit-helper-contract` (drawn track `bit-mechanisms`,
+chooser nonce `739c0098274f69cfc25fd435de0a325c`), refined to
+`R439-B509/source22/lower-frame/inverse-column-invariant`.
+Base commit; exact predecessor and dependency heads: base
+`ef5f39403610d231a0c27e8a6b4f08c10209bcd6`; predecessor [PR1368](https://github.com/CrocSwap/integer-mult-bounds/pull/1368) head
+`6ae784908c57f3aa5a2165d382b21a40accdf7a6` (r439-bblock-frame-barrier);
+second predecessor [PR1511](https://github.com/CrocSwap/integer-mult-bounds/pull/1511) head `f247c4e17b19e55c0db0ec8a1b660fd118adf014`
+(r439-source19-dual-frame-obstruction); frozen word
+`research/e8-bit-p111-tuned/tune/bitE8mut-p111-tuned.json.gz`, SHA-256
+`e3e301e62bc2b41dbf1ceac0acf3d35824ebeab1848c5a2142d4577b109da52f`.
+Concrete object and question: on the fixed eleven-wire F509 map of R439 phase-B
+block 509 (registers [17,22,33,38,50,60,67,77,87,92,468]), can any reversible
+XOR circuit with retained monotone common-frame operand promotion leave
+exceptional source 22 in a proper rational subframe of F509 and lower the
+complete paid-frame moment (0<a<1), including carriers, continuation and
+literal inverse cleanup? Or does the inverse-column frame invariant force
+every exit to F509?
+Scope status: refined (the drawn card's semantic-D H2/K question is already
+covered by [PR1194](https://github.com/CrocSwap/integer-mult-bounds/pull/1194)/PR1311 and adjacent work; this is the concrete open request
+in the handoffs thread https://github.com/CrocSwap/integer-mult-bounds/discussions/1166#discussioncomment-18911407
+and the open next question of [PR1511](https://github.com/CrocSwap/integer-mult-bounds/pull/1511)).
+Existing restriction being changed / claim audited: [PR1368](https://github.com/CrocSwap/integer-mult-bounds/pull/1368) leaves blocks 508
+and 509 open because of exceptional sources 19 and 22; [PR1511](https://github.com/CrocSwap/integer-mult-bounds/pull/1511) closes only
+B508/source 19 and names B509/source 22 as the remaining question.
+Method; explicit parameter ranges/candidate alphabet: exact rational span
+geometry over Q^8 and exact binary inversion over F2 on the pinned block map;
+all finite circuit lengths; no synthesis search. The all-length invariant of
+[PR1511](https://github.com/CrocSwap/integer-mult-bounds/pull/1511) audit.md Section 3 is applied to a separately pinned boundary.
+Nearest overlapping PRs and why this adds something: [PR1368](https://github.com/CrocSwap/integer-mult-bounds/pull/1368) (forward bounds,
+eight blocks; leaves 508/509 open), [PR1511](https://github.com/CrocSwap/integer-mult-bounds/pull/1511) (closes B508/source 19 only; open
+next question is exactly this). [Discussion 1476](https://github.com/CrocSwap/integer-mult-bounds/discussions/1476), [discussion 1503](https://github.com/CrocSwap/integer-mult-bounds/discussions/1503) and [PR1511](https://github.com/CrocSwap/integer-mult-bounds/pull/1511) claims concern
+helper-SWAP/three-cycle/dual-frame questions on other objects.
+Optimistic full cost budget; required observable improvement: no construction;
+acceptance is the fully priced obstruction itself (all eleven exits forced to
+F509, single-climb invoice minimal for 0<a<1, continuation and cleanup
+accounted), or an exact recorded obstruction if the invariant does not close.
+Success evidence; falsification test; time/compute cap: stdlib verifier with
+positive fixtures and failure controls (corrupt inverse bit, wrong
+orientation, singular map, frame descent, one-sided operand hosting);
+falsify by exhibiting a legal circuit or endpoint frame violating the claimed
+forced exits. One session; verification in the final quarter.
+Predecessor authors; intended artifact and reproduction command: Arham Tahir
+([PR1368](https://github.com/CrocSwap/integer-mult-bounds/pull/1368)), Andrey Mas ([PR1511](https://github.com/CrocSwap/integer-mult-bounds/pull/1511)); contributor Aaron Wang / navinor, with help
+from opencode. Artifact: research/r439-source22-dual-frame-obstruction/;
+reproduce `python3 -I -O -B research/r439-source22-dual-frame-obstruction/verify.py`.
@@ -0,0 +1 @@
+original/** -text -whitespace