diff --git a/src/fenigma/ocr.py b/src/fenigma/ocr.py
index ab9e1bd..0cfabf4 100644
--- a/src/fenigma/ocr.py
+++ b/src/fenigma/ocr.py
@@ -388,6 +388,98 @@ def parse_intel_blocks(text: str) -> list[dict]:
     return entries
 
 
+# Train-arrival intel is a different shape entirely, a station's absolute
+# grid ref, the rail's bearing from it, and a handful of waypoints given
+# only as a distance along that same bearing, plus a T=HH:MM:SS timestamp
+# on every line that this app makes no attempt to use, there's no game
+# clock to compare it against, so it's simply ignored:
+#   ARRIVAL STATION:
+#       Valle de Mula MainStation: J6 0:4
+#   Estimated arrival: T=10:16:50
+#   TRACK ALIGNMENT:
+#       Rail line runs straight. Bearing 090° from MainStation.
+#   FINAL APPROACH:
+#       Waypoint A - 6.00km from station: T=10:06:50
+#       Waypoint B - 4.00km from station: T=10:10:10
+# The station and each waypoint all come out RP-shaped (kind='rp'), same
+# as parse_intel_blocks()'s entries, so app.py can merge them the exact
+# same way, no separate code path needed on that side.
+_ARRIVAL_STATION_KEYWORD = "ARRIVAL STATION"
+_TRACK_ALIGNMENT_KEYWORD = "TRACK ALIGNMENT"
+_FINAL_APPROACH_KEYWORD = "FINAL APPROACH"
+
+_STATION_LINE_RE = re.compile(
+    rf"(\S+):\s*([A-T])\s*({_DIGIT_CLASS}{{1,2}})\s+({_DIGIT_CLASS})\s*[:;.,]\s*({_DIGIT_CLASS})"
+)
+_WAYPOINT_RE = re.compile(r"Waypoint\s+(\S+)\s*-\s*([\d.]+)\s*k?m\s+from\s+(\S+)", re.IGNORECASE)
+
+
+def _section(text: str, start_keyword: str, end_keywords: tuple[str, ...]) -> str | None:
+    """Text between a fuzzy match of `start_keyword` and whichever comes
+    first: a block-end '.' line, one of `end_keywords`, or the end of the
+    text. None if `start_keyword` isn't found at all. Scopes each
+    train-intel sub-parser to just its own section, so e.g. a
+    coincidental ':' + coord shape elsewhere in the text can't be
+    mistaken for the arrival station line."""
+    lines = text.splitlines()
+    start = next((i + 1 for i, line in enumerate(lines) if _fuzzy_contains(line.strip(), start_keyword)), None)
+    if start is None:
+        return None
+    collected = []
+    for line in lines[start:]:
+        stripped = line.strip()
+        if _BLOCK_END_RE.match(stripped) or any(_fuzzy_contains(stripped, k) for k in end_keywords):
+            break
+        collected.append(line)
+    return "\n".join(collected)
+
+
+def parse_train_intel(text: str) -> list[dict]:
+    """Parse a train-arrival intel block into RP-shaped entries: the
+    named station itself (absolute grid ref) plus one entry per waypoint,
+    each resolved via bearing (the rail's alignment) + distance (that
+    waypoint's distance along it) from the station, solve_location()'s
+    simplest case, one clue with both. Returns [] if this doesn't look
+    like a train intel block at all (no ARRIVAL STATION section)."""
+    station_block = _section(text, _ARRIVAL_STATION_KEYWORD, (_TRACK_ALIGNMENT_KEYWORD, _FINAL_APPROACH_KEYWORD))
+    if station_block is None:
+        return []
+    station_m = _STATION_LINE_RE.search(station_block)
+    if station_m is None:
+        return []
+    station_name, letter, y, x, yy = station_m.groups()
+    try:
+        station_coord = Coord(X=letter.upper(), Y=int(_fix_digits(y)), x=int(_fix_digits(x)), y=int(_fix_digits(yy)))
+    except ValueError:
+        return []
+
+    entries = [{
+        "kind": "rp", "name": station_name, "type_word": None, "id": None,
+        "raw": station_block.strip(), "clues": [], "coord": station_coord,
+    }]
+
+    track_block = _section(text, _TRACK_ALIGNMENT_KEYWORD, (_FINAL_APPROACH_KEYWORD,))
+    track_clues = _parse_all_clues(track_block) if track_block else []
+    bearing = next((c.bearing_deg for c in track_clues if c.bearing_deg is not None), None)
+    if bearing is None:
+        return entries  # the station alone is still useful even without the rail's bearing
+
+    approach_block = _section(text, _FINAL_APPROACH_KEYWORD, ())
+    if approach_block is None:
+        return entries
+
+    for m in _WAYPOINT_RE.finditer(approach_block):
+        letter_id, distance, ref_word = m.groups()
+        ref = station_name if ref_word.strip(".:").lower() == "station" else _clean_reference(ref_word)
+        entries.append({
+            "kind": "rp", "name": f"Waypoint {letter_id}", "type_word": None, "id": None,
+            "raw": m.group(0).strip(),
+            "clues": [Clue(reference=ref, bearing_deg=bearing, distance_km=float(_fix_digits(distance)))],
+            "coord": None,
+        })
+    return entries
+
+
 # Destruction reports are standalone one-liners, not tied to a block:
 #   "SupplyCache#2 Destroyed. Additional Requisition Granted."
 #   "Direct Hit! HostileTank#3 Destroyed."
@@ -466,6 +558,9 @@ def parse_text(text: str) -> ParsedInfo:
             continue
         info.targets[(target_type, entry["id"])] = (entry["raw"], entry["clues"], entry["coord"])
 
+    for entry in parse_train_intel(text):
+        info.reference_points[entry["name"]] = (entry["raw"], entry["clues"], entry["coord"])
+
     info.destroyed = parse_destroyed(text)
     return info