diff --git a/ROADMAP.md b/ROADMAP.md
index f4509df..5d83808 100644
--- a/ROADMAP.md
+++ b/ROADMAP.md
@@ -3,10 +3,10 @@
 Where DiffractoKit is going next, in order. `docs/parity.md` says what is held today and its known limits, and the GitHub
 issues track each piece of work. The numbers below are the ones `docs/parity.md` cites as "ROADMAP item n".
 
-1. **Diagnose before fixing.** Phase matching (`minimumVectors` = 2 makes "matrix" by exclusion; the cross-phase winner
-   ignores completeness), the ACOM mirror (a mirrored win carries `inPlaneAngle` + π with no flag; some off-grid
-   self-recoveries land 36 or 60 degrees away) and CIF symmetry (a group named by Hermann–Mauguin symbol alone, a partial
-   operation list). Each gets its cause stated first; what the diagnosis supports is fixed, the rest stays an open item.
+1. **Diagnose before fixing.** A number that reads wrong gets its cause stated first; what the diagnosis supports is
+   fixed, the rest stays an issue. Open today: the ACOM beam direction lands 3 to 6 degrees off the [001] and [011] zone
+   axes on the AlMgSi demo where excitation widths nearly tie (#65). The phase-matching, ACOM-mirror and CIF-symmetry
+   items of October are closed (#7, #8, #9). Listing on the Swift Package Index (#58) is open beside it.
 2. **Space-group facts, public.** Space group symbol, number and Hall symbol through spglib's definitions (no Wyckoff
    positions first); a public `ReflectionFamily` initialiser; `RingIndexing`'s "accepted" documented as a likelihood ratio,
    not family membership, or changed.
diff --git a/tools/beta-variants/README.md b/tools/beta-variants/README.md
index 43e5cc9..0cc14f5 100644
--- a/tools/beta-variants/README.md
+++ b/tools/beta-variants/README.md
@@ -1,8 +1,8 @@
-# beta'' variants: planted truth (DiffractoKit task 35)
+# beta'' variants: planted truth
 
 `planted.py` builds the 12 orientation-relationship variants of beta'' (Mg5Si6, C2/m) in Al as rotation matrices, with their
 pairwise disorientation angles in closed form and a set of seeded "measured" orientations whose nearest variant is known.
-numpy only; no orix or diffsims is run and no upstream number is claimed. The Mac pin script (kit audit step 0) adds
+numpy only; no orix or diffsims is run and no upstream number is claimed. `tools/pins/orix/pins.py` adds
 `Misorientation.equivalent()` / `reduce()` at atol 1e-4 from the same matrices.
 
 ```sh
@@ -19,15 +19,15 @@ Output: `variants.npy` (12, 3, 3), `measured.npy` (48, 3, 3), `truth.json`. Do n
 - 48 measured orientations (0.5 degree seeded noise, random 2/m setting) with the true variant; the nearest by disorientation
   is unique (second best at least 89 degrees away).
 
-## How the Mac session uses it
+## How the tests use it
 1. `CrystalPresets.betaDoublePrime` plus `OrientationRelationship.variants(matrix:precipitate:)` must return the 12 classes of
    `variants.npy` (compare as classes under the right group, atol 1e-9) and the angle matrix in `truth.json`.
 2. `nearestVariant(of:)` on `measured.npy` returns `truth.json.measured.true_variant` exactly.
-3. The pin script runs orix 0.15.0 `Misorientation.equivalent()` on the same R0 and compares at atol 1e-4 (the orix test's own).
+3. The orix pin script runs orix 0.15.0 `Misorientation.equivalent()` on the same R0 and compares at atol 1e-4 (the orix test's own).
 4. Frame: precipitate Cartesian e1 || a, e2 || b, e3 = e1 x e2; if the kit's `UnitCell` Cartesian setting differs, convert R0
    in the test wrapper and note it in `docs/parity.md`.
 
-## Open point for Paul
+## Open point
 The OR (a || [3 2 0]Al, b || [001]Al, c || [1 -3 0]Al) is derived in the script from the lattice parameters (105.25 degrees
-against beta = 105.3, a misfit 3.7 %); it is not copied from the paper. Scope question 1 (which OR the study used) decides it;
-replacing it means editing `R0` in `planted.py` only. Check against Andersen et al. 1998 before the figure quotes it.
+against beta = 105.3, a misfit 3.7 %); it is not copied from the paper. Replacing it means editing `R0` in `planted.py` only.
+Check against Andersen et al. 1998 before a figure quotes it.
diff --git a/tools/consumer-check/README.md b/tools/consumer-check/README.md
index f11a248..6cc7a8b 100644
--- a/tools/consumer-check/README.md
+++ b/tools/consumer-check/README.md
@@ -1,7 +1,7 @@
 # consumer-check
 
 A separate SwiftPM package that depends on DiffractoKit **by path** (`.package(path: "../..")`) and uses, from outside the
-package, the public API of slices 1 to 3: `UnitCell`, `MillerIndex`, `IntegerMatrix`, `SymmetryOperation`, `PointSymmetry`,
+package, the public API: `UnitCell`, `MillerIndex`, `IntegerMatrix`, `SymmetryOperation`, `PointSymmetry`,
 `LaueClass`, `CrystalSystem`, `Element`, `AtomSite`, `CrystalStructure`, `SymmetryDetection`, `CIFImport`, `CIFDocument`,
 `ReflectionList`, `Reflection`, `ReflectionFamily`, `ElectronScatteringFactors`, `StructureFactors`,
 `KinematicReflectionList`, and DiffractoGeometry's `ElectronWavelength`, `RingCalibration` (closed forms, `calibrate`,
@@ -15,26 +15,11 @@ a q at Au {111}'s g with σ = 1e-3 q accepted as {1 1 1}, β″'s [0 0 1] = (cos
 rendered by `GeometryChecks.swift` and recovered inside the apps' bands (radius 1 %, b/a 0.01, centre 1 px), and a 2:1
 ellipse's offset (4, 3) corrected to (2, 3)).
 
-DiffractoKit 0.1.0 adds `RingFitChecks.swift` and two sections of `main.swift`: `RingCalibration.q` / `radius`, `RingEvidence`
-(by hand and on a planted ring), `fitRing` / `calibrating` (equal to `fit`), `requireRingEvidence`, the three new refusals and
-`FitReport`'s new fields, `progress:` / `isCancelled:` on every fit, `LocalizedError` on the three error types, and the
-tetragonal settings. 397 checks at 0.0.4.
-
-The conventional setting and Niggli reduction add `ConventionalChecks.swift`: `UnitCell.niggliReduced` / `ReducedCell` (fcc gold's primitive cell, a skewed basis of
-it, an invalid cell), `UnitCell.conventionalCell` / `ConventionalCell` / `BravaisLattice` (gold primitive -> cF with 4 centring
-translations, bcc -> cI, a C-centred monoclinic primitive cell -> mC with b the unique axis), `CrystalStructure.setting` /
-`conventionalSetting` / `ConventionalSetting` and its four conversions (gold: 192 operations in the conventional cell and 48 in
-the primitive file cell; rutile with its 4-fold axis along a: 4/mmm, 16 operations, the file's (1 0 0) the conventional
-(0 0 1)), and `detectingSymmetry(conventional:tolerance:search:)` with `SymmetryDetection.Tolerance` / `Search` (a 0.012 Å
-displacement in a 30 Å cell: symmetric at the fractional 5e-3, not at the Cartesian 1e-3 Å; the cell list equals the brute-force
-scan). 530 checks in all.
-
-DiffractoKit 0.2.0 adds `PolicyChecks.swift`: every `Policy` type (`ReflectionList`, `CrystalStructure.ValidationPolicy`,
-`SymmetryDetection`, `CIFImport`, `EllipseCalibration`, `RingEvidence`, `RingCalibration`) with its `.default` read field by
-field, one non-default use whose outcome a reader verifies by hand (a 0.3 Å contact usable under a 0.2 Å limit, a 120 degree
-gap fitted under a coverage bound of 20 bins, occupancies 0.50 and 0.51 two species at tolerance 0), the refusals of an invalid
-value (a position tolerance outside its domain, an `Int.max` bin count, a NaN residual cap), `SymmetryOperation(xyz:maximumMagnitude:)`,
-`RingIndexing`'s `minimumMargin:` and `reach`, and `retainingAll:` (`kept` and `families` equal, `all` empty).
+`RingFitChecks.swift` covers the ring fit (`RingCalibration.q` / `radius`, `RingEvidence`, `fitRing` / `calibrating`, the
+refusals, `FitReport`, `progress:` / `isCancelled:`, `LocalizedError` on the error types, the tetragonal settings);
+`ConventionalChecks.swift` the Niggli reduction, the conventional cell and `SymmetryDetection` with its tolerance and search;
+`PolicyChecks.swift` every `Policy` type read field by field, one non-default use each and the refusals of invalid values.
+About 530 checks in all.
 
 ```
 tools/consumer-check/run.sh