diff --git a/context/open-questions.md b/context/open-questions.md
index da52e35..559981b 100644
--- a/context/open-questions.md
+++ b/context/open-questions.md
@@ -1357,3 +1357,15 @@ EXPERT-only; the player+0x260 WRITER has never been read).  Settled/confirmed [w
   expert-only -> best read: a genuine 4.0->4.10 balance change (his hands-on spans the earlier
   machine).  Tester-notes line warranted (veteran heat is not a bug).  The vet-vs-expert heat
   MAGNITUDE coefficient stays open [T4].
+- **NARC/STREAK per-type seeker maneuverability (Oracle: the "angry swarm of bees" at high
+  aspect angles near a maneuvering target)** [T3→T4]: the port's steering gains (4.0 turn /
+  8.0 loft) are PORT-TUNED constants applied uniformly to every missile family; Oracle's
+  memory says the maneuverable types behaved differently.  Candidate: an authored per-model
+  turn-rate field in the missile model record (we consume +0x44 maxTimeOfFlight, +0x48
+  thrust, +0x50 detonationMode — +0x4c is UNREAD).  Read +0x4c across the 8 cached missile
+  models; if it differentiates NARC/STREAK, wire it as the turn gain.  Post-Friday.
+- **Gauss carrier confirmed vs content (Oracle: "the Mischief")**: Mischief = lok2 (the exe's
+  variant-name table: vul1="Mad Dog", lok2="Mischief", mad2="Zanin Neko"), and the Loki
+  family's L4GAUGE.CFG section authors GAUSS + AmmoBinGAUSS gauge rows — consistent.  (A
+  separate section authors GAUSS_1/GAUSS_2 twin rows at letters H/O — which family that is
+  remains unidentified; low priority.)