diff --git a/README.md b/README.md
index 78093ce..b7f6c66 100644
--- a/README.md
+++ b/README.md
@@ -35,9 +35,9 @@ infrastructure, splits, logging, versioning) stays yours.
 
 **1. Raw events.** Any PyTorch `Dataset` yielding
 `raw[i] -> {"event_no": int, "pulses": [P, F] float32, "sensor_key": [P] int}`
-(stated canonically in `spine/data/datamodule.py`). Pulses stay raw: the
-sampler builds its cutoffs and dt targets in detector units, and
-standardization happens later at collate. Columns follow the task's
+(stated canonically in `spine/data/datamodule.py`). Pulses stay raw: pretext
+tasks make their sampling decisions and build their targets in detector
+units, and standardization happens later at collate. Columns follow the task's
 `FeatureLayout`, by default `(x, y, z, t, charge)`; pass a different layout
 instead of reordering your data. `sensor_key` is a unique integer identity
 per sensor carried in the data itself; single PMT detectors can use any
@@ -46,22 +46,24 @@ stable per sensor id. Reference readers:
 `SQLiteDataset` / `LMDBDataset` adapted via
 `spine_graphnet.readers.GraphNetRawDataset` (see `examples/graphnet_demo.py`).
 
-**2. A geometry asset.** An `.npz` with per sensor arrays: `xyz [S, 3]` in
-the same units as the pulse coordinates, `knn_idx [S, K]` neighbours sorted
-by distance (the sampler walks it to the nearest dark sensor, so K must be
-large enough that one is always found; the full sorted list is safest), and
-one unique integer key array whose name you pass to
-`load_geometry(path, sensor_key=...)`. The reader's `sensor_key` values must
-resolve through exactly these keys; coordinate matching is deliberately not
-supported. `build_geometry_asset` in `examples/graphnet_demo.py` shows a
-build from a pulse file.
+**2. A geometry asset.** An `.npz` with per sensor arrays: always
+`xyz [S, 3]` in the same units as the pulse coordinates, and one unique
+integer key array whose name you pass to `load_geometry(path, sensor_key=...)`.
+A task can require more; CURTAIN for example needs `knn_idx [S, K]`,
+neighbours sorted by distance, to anchor its negatives at the nearest dark
+sensor (K large enough that one is always found; the full sorted list is
+safest). The reader's `sensor_key` values must resolve through exactly these
+keys; coordinate matching is deliberately not supported.
+`build_geometry_asset` in `examples/graphnet_demo.py` shows a build from a
+pulse file.
 
 **3. Selections.** SPINE owns no split: `fit` takes separate train and val
-Datasets and you keep them disjoint. Every selected event must be guaranteed
-splittable: pre filter with `spine.pretext.curtain.sampler.can_always_split`
-using the same `min_visible`/`min_future` you give the task, and float32
-times. `make_sample` raises on events that slip through rather than skipping
-them silently.
+Datasets and you keep them disjoint. Every selected event must satisfy the
+task's sampling requirements; tasks raise on events that fall short instead
+of skipping them silently, so pre filter your selection with the task's own
+predicate. For CURTAIN that is
+`spine.pretext.curtain.sampler.can_always_split`, called with the same
+`min_visible`/`min_future` you give the task and float32 times.
 
 **4. Feature scaling.** A `FeatureScaler` subclass (`scale_pulses` and
 `scale_positions`, both applying the same xyz factors) for your detector, or