diff --git a/bincio/extract/cli.py b/bincio/extract/cli.py
index d57b03d..814316e 100644
--- a/bincio/extract/cli.py
+++ b/bincio/extract/cli.py
@@ -1,6 +1,7 @@
 """bincio extract — CLI command."""
 
 import json
+import os
 import sys
 from concurrent.futures import ProcessPoolExecutor, as_completed
 from pathlib import Path
@@ -12,15 +13,88 @@ from rich.progress import BarColumn, MofNCompleteColumn, Progress, TextColumn, T
 
 from bincio.extract.config import ExtractConfig, default_config, load_config
 from bincio.extract.dedup import ActivityRecord, DedupIndex
-from bincio.extract.metrics import compute
-from bincio.extract.models import ParsedActivity
-from bincio.extract.parsers.factory import is_supported, parse_file
-from bincio.extract.strava_csv import StravaMetadata
-from bincio.extract.writer import build_summary, make_activity_id, write_activity, write_index
+from bincio.extract.parsers.factory import is_supported
 
 console = Console()
 
 
+# ── per-worker state (set once via initializer, never re-pickled) ─────────────
+
+_known_hashes: frozenset = frozenset()
+_strava_lookup: dict = {}
+_output_dir: Path = Path(".")
+_privacy: str = "public"
+_rdp_epsilon: float = 0.0001
+
+
+def _worker_init(
+    known_hashes: frozenset,
+    strava_lookup: dict,
+    output_dir: Path,
+    privacy: str,
+    rdp_epsilon: float,
+) -> None:
+    global _known_hashes, _strava_lookup, _output_dir, _privacy, _rdp_epsilon
+    _known_hashes = known_hashes
+    _strava_lookup = strava_lookup
+    _output_dir = output_dir
+    _privacy = privacy
+    _rdp_epsilon = rdp_epsilon
+
+
+def _process_file(path: Path) -> dict:
+    """Runs inside a worker process. Only receives a Path (tiny pickle).
+    All heavy shared data (_known_hashes, _strava_lookup, etc.) is already
+    in the worker's memory from the initializer — zero per-task overhead.
+    """
+    from bincio.extract.metrics import compute
+    from bincio.extract.parsers.factory import parse_file
+    from bincio.extract.writer import build_summary, make_activity_id, write_activity
+
+    try:
+        activity = parse_file(path)
+    except Exception as exc:
+        return {"status": "error", "path": str(path), "error": str(exc)}
+
+    # Exact-duplicate check (free — just a set lookup)
+    if activity.source_hash in _known_hashes:
+        return {"status": "duplicate"}
+
+    # Enrich from Strava CSV
+    row = _strava_lookup.get(activity.source_file)
+    if row:
+        if not activity.title:
+            activity.title = row.get("Activity Name", "").strip() or None
+        if not activity.description:
+            activity.description = row.get("Activity Description", "").strip() or None
+        if not activity.strava_id:
+            activity.strava_id = row.get("Activity ID", "").strip() or None
+
+    try:
+        metrics = compute(activity)
+        activity_id = make_activity_id(activity)
+        write_activity(
+            activity, metrics, _output_dir,
+            privacy=_privacy,
+            rdp_epsilon=_rdp_epsilon,
+        )
+        summary = build_summary(activity, metrics, activity_id, _privacy)
+    except Exception as exc:
+        return {"status": "error", "path": str(path), "error": str(exc)}
+
+    return {
+        "status": "ok",
+        "summary": summary,
+        "id": activity_id,
+        "hash": activity.source_hash,
+        "started_at": activity.started_at.isoformat(),
+        "distance_m": metrics.distance_m,
+        "source": summary.get("source"),
+    }
+
+
+# ── CLI ────────────────────────────────────────────────────────────────────────
+
 @click.command()
 @click.option("--config", "config_path", type=click.Path(exists=True), default=None,
               help="Path to extract_config.yaml (default: ./extract_config.yaml).")
@@ -32,49 +106,48 @@ console = Console()
               help="Process a single file and print JSON to stdout.")
 @click.option("--since", default=None, metavar="YYYY-MM-DD",
               help="Only process files modified after this date.")
-@click.option("--workers", default=4, show_default=True,
-              help="Number of parallel worker processes.")
+@click.option("--workers", default=None, type=int,
+              help="Parallel worker processes (default: CPU count).")
 def extract(
     config_path: Optional[str],
     input_dir: Optional[str],
     output_dir: Optional[str],
     single_file: Optional[str],
     since: Optional[str],
-    workers: int,
+    workers: Optional[int],
 ) -> None:
     """Parse GPX/FIT/TCX files and write BAS JSON data store."""
 
-    # ── single file mode ─────────────────────────────────────────────────────
     if single_file:
         _process_single(Path(single_file))
         return
 
-    # ── load config ──────────────────────────────────────────────────────────
     cfg = _resolve_config(config_path, input_dir, output_dir)
     cfg.output_dir.mkdir(parents=True, exist_ok=True)
 
-    # ── gather files ─────────────────────────────────────────────────────────
     files = _collect_files(cfg, since)
     if not files:
         console.print("[yellow]No supported files found.[/yellow]")
         return
     console.print(f"Found [bold]{len(files)}[/bold] activity files.")
 
-    # ── Strava metadata ──────────────────────────────────────────────────────
-    strava_meta: Optional[StravaMetadata] = None
+    # Build strava lookup once (serialised dict, sent to workers via initializer)
+    strava_lookup: dict = {}
     if cfg.metadata_csv and cfg.metadata_csv.exists():
-        strava_meta = StravaMetadata(cfg.metadata_csv)
+        from bincio.extract.strava_csv import StravaMetadata
+        strava_lookup = StravaMetadata(cfg.metadata_csv)._by_filename
         console.print(f"Loaded Strava metadata from [cyan]{cfg.metadata_csv.name}[/cyan].")
 
-    # ── dedup index ──────────────────────────────────────────────────────────
     dedup = DedupIndex(output_dir=cfg.output_dir)
+    known_hashes: frozenset = frozenset(dedup._by_hash.keys())
 
-    # ── process ──────────────────────────────────────────────────────────────
-    summaries: list[dict] = []
-    errors: list[tuple[Path, str]] = []
-    skipped = 0
+    n_workers = workers or os.cpu_count() or 4
+    console.print(f"Using [bold]{n_workers}[/bold] worker processes.")
 
     owner = {"handle": cfg.owner_handle, "display_name": cfg.owner_display_name}
+    summaries: list[dict] = []
+    errors: list[tuple[str, str]] = []
+    skipped = 0
 
     with Progress(
         TextColumn("[progress.description]{task.description}"),
@@ -85,80 +158,52 @@ def extract(
     ) as progress:
         task = progress.add_task("Processing...", total=len(files))
 
-        with ProcessPoolExecutor(max_workers=workers) as pool:
-            futures = {pool.submit(_parse_worker, f): f for f in files}
+        with ProcessPoolExecutor(
+            max_workers=n_workers,
+            initializer=_worker_init,
+            initargs=(known_hashes, strava_lookup, cfg.output_dir, cfg.default_privacy, cfg.track.rdp_epsilon),
+        ) as pool:
+            futures = {pool.submit(_process_file, f): f for f in files}
             for future in as_completed(futures):
-                path = futures[future]
                 progress.advance(task)
-                try:
-                    activity = future.result()
-                except Exception as exc:
-                    errors.append((path, str(exc)))
-                    continue
+                result = future.result()
 
-                # ── incremental skip ──────────────────────────────────────
-                if cfg.incremental:
-                    existing_id = dedup.is_exact_duplicate(activity.source_hash)
-                    if existing_id:
-                        skipped += 1
-                        continue
+                if result["status"] == "duplicate":
+                    skipped += 1
+                elif result["status"] == "error":
+                    errors.append((result["path"], result["error"]))
+                else:
+                    # Near-duplicate check — must be sequential (stateful)
+                    from datetime import datetime
+                    started_at = datetime.fromisoformat(result["started_at"])
+                    near_id = dedup.find_near_duplicate(started_at, result["distance_m"])
 
-                # ── enrich from Strava CSV ────────────────────────────────
-                if strava_meta:
-                    strava_meta.enrich(activity.source_file, activity)
+                    if near_id:
+                        canonical = dedup.pick_canonical(near_id, result.get("source"))
+                        if canonical != "__new__":
+                            _patch_duplicate_of(cfg.output_dir, result["id"], near_id)
+                            skipped += 1
+                            continue
+                        _patch_duplicate_of(cfg.output_dir, near_id, result["id"])
+                        dedup._records[near_id].duplicate_of = result["id"]
 
-                # ── compute metrics ───────────────────────────────────────
-                metrics = compute(activity)
+                    dedup.register(ActivityRecord(
+                        id=result["id"],
+                        source_hash=result["hash"],
+                        started_at=started_at,
+                        distance_m=result["distance_m"],
+                        source=result.get("source"),
+                    ))
+                    summaries.append(result["summary"])
 
-                # ── deduplication ─────────────────────────────────────────
-                activity_id = make_activity_id(activity)
-                duplicate_of: Optional[str] = None
-
-                near_dup_id = dedup.find_near_duplicate(
-                    activity.started_at, metrics.distance_m
-                )
-                if near_dup_id:
-                    source = _infer_source(activity)
-                    canonical = dedup.pick_canonical(near_dup_id, source)
-                    if canonical == "__new__":
-                        # New one is better — mark existing as duplicate
-                        existing = dedup._records[near_dup_id]
-                        existing.duplicate_of = activity_id
-                    else:
-                        duplicate_of = near_dup_id
-
-                # ── write files ───────────────────────────────────────────
-                written_id = write_activity(
-                    activity, metrics, cfg.output_dir,
-                    privacy=cfg.default_privacy,
-                    duplicate_of=duplicate_of,
-                    rdp_epsilon=cfg.track.rdp_epsilon,
-                )
-
-                # Register in dedup index
-                dedup.register(ActivityRecord(
-                    id=written_id,
-                    source_hash=activity.source_hash,
-                    started_at=activity.started_at,
-                    distance_m=metrics.distance_m,
-                    source=_infer_source(activity),
-                ))
-
-                if duplicate_of is None:
-                    summaries.append(
-                        build_summary(activity, metrics, written_id, cfg.default_privacy)
-                    )
-
-    # ── write index.json ──────────────────────────────────────────────────────
-    # Merge with any existing summaries from previous incremental runs
-    existing_index = _load_existing_summaries(cfg.output_dir)
-    all_summaries = {s["id"]: s for s in existing_index}
+    from bincio.extract.writer import write_index
+    existing = _load_existing_summaries(cfg.output_dir)
+    merged = {s["id"]: s for s in existing}
     for s in summaries:
-        all_summaries[s["id"]] = s
-    write_index(list(all_summaries.values()), cfg.output_dir, owner)
+        merged[s["id"]] = s
+    write_index(list(merged.values()), cfg.output_dir, owner)
     dedup.save()
 
-    # ── summary ───────────────────────────────────────────────────────────────
     console.print(
         f"\n[green]Done.[/green] "
         f"Processed [bold]{len(summaries)}[/bold] activities, "
@@ -168,28 +213,22 @@ def extract(
     if errors:
         console.print("\n[red]Errors:[/red]")
         for path, msg in errors[:20]:
-            console.print(f"  {path.name}: {msg}")
+            console.print(f"  {Path(path).name}: {msg}")
         if len(errors) > 20:
             console.print(f"  ... and {len(errors) - 20} more.")
 
 
 # ── helpers ───────────────────────────────────────────────────────────────────
 
-def _parse_worker(path: Path) -> ParsedActivity:
-    """Run in worker process — imports are isolated."""
-    from bincio.extract.parsers.factory import parse_file
-    return parse_file(path)
-
-
 def _process_single(path: Path) -> None:
+    from bincio.extract.metrics import compute
     from bincio.extract.parsers.factory import parse_file
+    from bincio.extract.writer import build_summary, make_activity_id
     try:
         activity = parse_file(path)
         metrics = compute(activity)
         activity_id = make_activity_id(activity)
-        from bincio.extract.writer import build_summary
-        result = build_summary(activity, metrics, activity_id)
-        click.echo(json.dumps(result, indent=2))
+        click.echo(json.dumps(build_summary(activity, metrics, activity_id), indent=2))
     except Exception as exc:
         console.print(f"[red]Error:[/red] {exc}")
         sys.exit(1)
@@ -221,51 +260,39 @@ def _resolve_config(
 
 
 def _collect_files(cfg: ExtractConfig, since: Optional[str]) -> list[Path]:
-    from bincio.extract.parsers.factory import is_supported
-    import os
     from datetime import datetime
-
     since_ts: Optional[float] = None
     if since:
         since_ts = datetime.strptime(since, "%Y-%m-%d").timestamp()
-
     files = []
     for d in cfg.input_dirs:
         if not d.exists():
             console.print(f"[yellow]Warning:[/yellow] input dir not found: {d}")
             continue
         for path in d.rglob("*"):
-            if not path.is_file():
-                continue
-            if not is_supported(path):
-                continue
-            if since_ts and path.stat().st_mtime < since_ts:
-                continue
-            files.append(path)
+            if path.is_file() and is_supported(path):
+                if not since_ts or path.stat().st_mtime >= since_ts:
+                    files.append(path)
     return files
 
 
 def _load_existing_summaries(output_dir: Path) -> list[dict]:
-    index_path = output_dir / "index.json"
-    if not index_path.exists():
+    p = output_dir / "index.json"
+    if not p.exists():
         return []
     try:
-        data = json.loads(index_path.read_text())
-        return data.get("activities", [])
+        return json.loads(p.read_text()).get("activities", [])
     except Exception:
         return []
 
 
-def _infer_source(activity: ParsedActivity) -> Optional[str]:
-    if activity.strava_id:
-        return "strava_export"
-    name = activity.source_file.lower()
-    if "activity" in name and len(name.split(".")) >= 3:
-        return "karoo"
-    if name.endswith((".fit", ".fit.gz")):
-        return "fit_file"
-    if name.endswith((".gpx", ".gpx.gz")):
-        return "gpx_file"
-    if name.endswith((".tcx", ".tcx.gz")):
-        return "tcx_file"
-    return None
+def _patch_duplicate_of(output_dir: Path, activity_id: str, canonical_id: str) -> None:
+    p = output_dir / "activities" / f"{activity_id}.json"
+    if not p.exists():
+        return
+    try:
+        data = json.loads(p.read_text())
+        data["duplicate_of"] = canonical_id
+        p.write_text(json.dumps(data, indent=2, ensure_ascii=False))
+    except Exception:
+        pass
diff --git a/bincio/extract/dedup.py b/bincio/extract/dedup.py
index 642b3f9..90b50d6 100644
--- a/bincio/extract/dedup.py
+++ b/bincio/extract/dedup.py
@@ -108,6 +108,8 @@ class DedupIndex:
             if distance_m is None or r.distance_m is None:
                 continue
             ref = max(distance_m, r.distance_m)
+            if ref < 1.0:
+                continue  # both near-zero (indoor/manual) — skip distance check
             if abs(distance_m - r.distance_m) / ref < 0.05:
                 return r.id
         return None
diff --git a/bincio/extract/metrics.py b/bincio/extract/metrics.py
index ad589f8..f492825 100644
--- a/bincio/extract/metrics.py
+++ b/bincio/extract/metrics.py
@@ -1,19 +1,28 @@
 """Compute aggregated metrics from a ParsedActivity.
 
 All calculations are self-contained — no external state needed.
+Uses inline haversine rather than geopy.geodesic to keep the hot path fast.
 """
 
 import math
 from dataclasses import dataclass
-from datetime import datetime
 from typing import Optional
 
-from geopy.distance import geodesic
-
 from bincio.extract.models import DataPoint, ParsedActivity
 
 # Speed below which we consider the athlete stopped (km/h)
 _STOPPED_THRESHOLD_KMH = 1.0
+_EARTH_R = 6_371_000.0  # metres
+
+
+def _haversine_m(lat1: float, lon1: float, lat2: float, lon2: float) -> float:
+    """Great-circle distance in metres. ~10x faster than geopy.geodesic."""
+    phi1 = math.radians(lat1)
+    phi2 = math.radians(lat2)
+    dphi = phi2 - phi1
+    dlam = math.radians(lon2 - lon1)
+    a = math.sin(dphi * 0.5) ** 2 + math.cos(phi1) * math.cos(phi2) * math.sin(dlam * 0.5) ** 2
+    return 2.0 * _EARTH_R * math.asin(math.sqrt(min(a, 1.0)))
 
 
 @dataclass
@@ -30,7 +39,7 @@ class ComputedMetrics:
     avg_cadence_rpm: Optional[int]
     avg_power_w: Optional[int]
     max_power_w: Optional[int]
-    bbox: Optional[tuple[float, float, float, float]]  # min_lon, min_lat, max_lon, max_lat
+    bbox: Optional[tuple[float, float, float, float]]   # min_lon, min_lat, max_lon, max_lat
     start_latlng: Optional[tuple[float, float]]
     end_latlng: Optional[tuple[float, float]]
 
@@ -41,10 +50,8 @@ def compute(activity: ParsedActivity) -> ComputedMetrics:
         return _empty()
 
     duration_s = _duration(pts)
-    distance_m = _distance(pts)
-    moving_time_s, moving_speed_kmh = _moving_stats(pts)
+    distance_m, moving_time_s, avg_speed_kmh, max_speed_kmh = _gps_stats(pts)
     gain, loss = _elevation(pts)
-    max_speed = _max_speed(pts)
     avg_hr, max_hr = _hr_stats(pts)
     avg_cad = _avg_nonnull([p.cadence_rpm for p in pts])
     avg_pow = _avg_nonnull([p.power_w for p in pts])
@@ -58,8 +65,8 @@ def compute(activity: ParsedActivity) -> ComputedMetrics:
         moving_time_s=moving_time_s,
         elevation_gain_m=round(gain, 1) if gain is not None else None,
         elevation_loss_m=round(abs(loss), 1) if loss is not None else None,
-        avg_speed_kmh=round(moving_speed_kmh, 2) if moving_speed_kmh else None,
-        max_speed_kmh=round(max_speed, 2) if max_speed else None,
+        avg_speed_kmh=round(avg_speed_kmh, 2) if avg_speed_kmh else None,
+        max_speed_kmh=round(max_speed_kmh, 2) if max_speed_kmh else None,
         avg_hr_bpm=avg_hr,
         max_hr_bpm=max_hr,
         avg_cadence_rpm=avg_cad,
@@ -71,66 +78,75 @@ def compute(activity: ParsedActivity) -> ComputedMetrics:
     )
 
 
-# ── helpers ──────────────────────────────────────────────────────────────────
+# ── single-pass GPS stats ──────────────────────────────────────────────────────
+# distance, moving time, avg speed, and max speed are all derived from the same
+# per-segment loop, so we compute them in one pass instead of four.
 
-def _duration(pts: list[DataPoint]) -> Optional[int]:
-    if len(pts) < 2:
-        return None
-    return int((pts[-1].timestamp - pts[0].timestamp).total_seconds())
+def _gps_stats(
+    pts: list[DataPoint],
+) -> tuple[Optional[float], Optional[int], Optional[float], Optional[float]]:
+    """Return (distance_m, moving_time_s, avg_speed_kmh, max_speed_kmh)."""
 
-
-def _distance(pts: list[DataPoint]) -> Optional[float]:
-    """Prefer device-recorded cumulative distance; fall back to GPS geodesic."""
-    # If the last point has a device distance, use it
-    last_dist = next(
+    # Prefer device-recorded cumulative distance (FIT files always have this)
+    device_dist = next(
         (p.distance_m for p in reversed(pts) if p.distance_m is not None), None
     )
-    if last_dist is not None:
-        return round(last_dist, 1)
 
-    # GPS fallback
-    total = 0.0
-    has_gps = False
-    for a, b in zip(pts, pts[1:]):
-        if a.lat is None or a.lon is None or b.lat is None or b.lon is None:
-            continue
-        has_gps = True
-        total += geodesic((a.lat, a.lon), (b.lat, b.lon)).meters
-    return round(total, 1) if has_gps else None
-
-
-def _moving_stats(pts: list[DataPoint]) -> tuple[Optional[int], Optional[float]]:
-    """Return (moving_time_s, avg_speed_kmh_over_moving_time)."""
     moving_s = 0
     moving_dist_m = 0.0
-    has_gps = False
+    total_dist_m = 0.0
+    max_seg_kmh = 0.0
+    has_data = False
+
+    # Device speed values (used for max if present)
+    device_max_kmh: Optional[float] = None
+    if any(p.speed_kmh is not None for p in pts):
+        device_max_kmh = max(p.speed_kmh for p in pts if p.speed_kmh is not None)
 
     for a, b in zip(pts, pts[1:]):
         dt = (b.timestamp - a.timestamp).total_seconds()
         if dt <= 0:
             continue
 
-        # Compute speed for this interval from GPS
         if a.lat is not None and a.lon is not None and b.lat is not None and b.lon is not None:
-            has_gps = True
-            seg_m = geodesic((a.lat, a.lon), (b.lat, b.lon)).meters
+            seg_m = _haversine_m(a.lat, a.lon, b.lat, b.lon)
             seg_kmh = (seg_m / dt) * 3.6
+            has_data = True
         elif a.speed_kmh is not None:
             seg_kmh = a.speed_kmh
             seg_m = (seg_kmh / 3.6) * dt
-            has_gps = True  # speed data present
+            has_data = True
         else:
             continue
 
+        total_dist_m += seg_m
+        if seg_kmh > max_seg_kmh:
+            max_seg_kmh = seg_kmh
+
         if seg_kmh >= _STOPPED_THRESHOLD_KMH:
             moving_s += int(dt)
             moving_dist_m += seg_m
 
-    if not has_gps or moving_s == 0:
-        return None, None
+    if not has_data:
+        return device_dist, None, None, None
 
-    avg_kmh = (moving_dist_m / moving_s) * 3.6
-    return moving_s, avg_kmh
+    distance_m = device_dist if device_dist is not None else round(total_dist_m, 1)
+    moving_time_s = moving_s if moving_s > 0 else None
+    avg_speed_kmh = (moving_dist_m / moving_s) * 3.6 if moving_s > 0 else None
+    # Prefer device speed for max (more stable than GPS-derived per-second spikes)
+    max_speed_kmh = device_max_kmh if device_max_kmh is not None else (
+        max_seg_kmh if max_seg_kmh > 0 else None
+    )
+
+    return distance_m, moving_time_s, avg_speed_kmh, max_speed_kmh
+
+
+# ── remaining helpers ──────────────────────────────────────────────────────────
+
+def _duration(pts: list[DataPoint]) -> Optional[int]:
+    if len(pts) < 2:
+        return None
+    return int((pts[-1].timestamp - pts[0].timestamp).total_seconds())
 
 
 def _elevation(pts: list[DataPoint]) -> tuple[Optional[float], Optional[float]]:
@@ -147,24 +163,6 @@ def _elevation(pts: list[DataPoint]) -> tuple[Optional[float], Optional[float]]:
     return gain, loss
 
 
-def _max_speed(pts: list[DataPoint]) -> Optional[float]:
-    # Prefer device speed; fall back to GPS-derived
-    device_speeds = [p.speed_kmh for p in pts if p.speed_kmh is not None]
-    if device_speeds:
-        return max(device_speeds)
-    # GPS-derived max
-    gps_speeds = []
-    for a, b in zip(pts, pts[1:]):
-        if a.lat is None or b.lat is None:
-            continue
-        dt = (b.timestamp - a.timestamp).total_seconds()
-        if dt <= 0:
-            continue
-        m = geodesic((a.lat, a.lon), (b.lat, b.lon)).meters
-        gps_speeds.append((m / dt) * 3.6)
-    return max(gps_speeds) if gps_speeds else None
-
-
 def _hr_stats(pts: list[DataPoint]) -> tuple[Optional[int], Optional[int]]:
     hrs = [p.hr_bpm for p in pts if p.hr_bpm is not None]
     if not hrs:
diff --git a/bincio/extract/parsers/tcx.py b/bincio/extract/parsers/tcx.py
index bda8998..c80e9d1 100644
--- a/bincio/extract/parsers/tcx.py
+++ b/bincio/extract/parsers/tcx.py
@@ -8,18 +8,24 @@ from lxml import etree
 from bincio.extract.models import DataPoint, ParsedActivity
 from bincio.extract.sport import normalise_sport
 
-_NS = {
+_NS_HTTP  = {
     "tcx": "http://www.garmin.com/xmlschemas/TrainingCenterDatabase/v2",
     "ext": "http://www.garmin.com/xmlschemas/ActivityExtension/v2",
 }
+_NS_HTTPS = {
+    "tcx": "https://www.garmin.com/xmlschemas/TrainingCenterDatabase/v2",
+    "ext": "https://www.garmin.com/xmlschemas/ActivityExtension/v2",
+}
 
 
 class TcxParser:
     def parse(self, path: Path, raw_bytes: bytes) -> ParsedActivity:
-        # Some exporters (e.g. Garmin) prepend whitespace before the XML
-        # declaration, which is technically invalid. Strip it.
+        # Some exporters prepend whitespace before the XML declaration. Strip it.
         root = etree.fromstring(raw_bytes.lstrip())
 
+        # Garmin sometimes uses https:// instead of http:// in the namespace URI.
+        _NS = _NS_HTTPS if b"https://www.garmin.com" in raw_bytes else _NS_HTTP
+
         activities = root.findall(".//tcx:Activity", _NS)
         if not activities:
             raise ValueError(f"No Activity elements found in {path.name}")
diff --git a/bincio/extract/writer.py b/bincio/extract/writer.py
index 8872011..4376480 100644
--- a/bincio/extract/writer.py
+++ b/bincio/extract/writer.py
@@ -12,11 +12,13 @@ from bincio.extract.timeseries import build_timeseries
 
 
 def make_activity_id(activity: ParsedActivity) -> str:
-    """Generate a BAS activity ID from started_at + optional title slug."""
-    ts = activity.started_at
-    # Compact ISO format: 2024-06-01T073012+0200
-    tz_str = ts.strftime("%z")  # e.g. "+0200" or ""
-    ts_part = ts.strftime("%Y-%m-%dT%H%M%S") + (tz_str or "Z")
+    """Generate a BAS activity ID from started_at + optional title slug.
+
+    Always uses UTC with Z suffix so IDs are URL-safe (no + chars).
+    """
+    from datetime import timezone
+    ts = activity.started_at.astimezone(timezone.utc)
+    ts_part = ts.strftime("%Y-%m-%dT%H%M%SZ")
 
     if activity.title:
         slug = _slugify(activity.title)
diff --git a/pyproject.toml b/pyproject.toml
index 66a7e88..f6f8893 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -19,7 +19,6 @@ dependencies = [
     # Data
     "pandas>=2.2",
     # Geo
-    "geopy>=2.4",
     "rdp>=0.8",
     # Config & CLI
     "pyyaml>=6.0",