Known limitations

Scan capture (Live Scan & AR Trace)

PackPilot offers two scan methods: Live Scan, which captures a space's interior volume from the device's live scene depth as you move the phone around inside, and AR Trace, the manual alternative where you tap the floor corners and extrude to a box. The notes below apply to both.

• Captures are bounding boxes, not meshes. A cylinder is captured as the box that contains it; a soft bag is captured at whatever size it's currently bulged to. Trim oversized scans in the confirmation sheet before saving. Choosing the right Shape Type is what changes how the solver treats the item — the scan view doesn't yet visualise that shape change.
• Featureless and dim rooms struggle. ARKit needs visual texture to lock on. Plain carpet with no edges, blank walls, or a very dim room can prevent the first plane detection. The coaching overlay nudges you toward better surfaces.
• Tracking can drift over long camera motion. Pointing at a featureless wall for several seconds degrades ARKit's world tracking, and the wireframe may shimmer or slide. Re-align recovers it; the app doesn't detect drift automatically yet.
• No scanning on unsupported devices. Both Live Scan and AR Trace need an ARKit-compatible device, and Live Scan additionally relies on scene depth sensing. Where they aren't available, every flow falls back to manual entry, which is fully functional — just not visually guided.

AR Placement Guide

• Corner order matters. You align by tapping the front-left then front-right base corner. Tap them in the other order and the depth axis flips, so ghosts point the wrong way. Re-align from the toolbar menu fixes it.
• Each session starts cold. There's no persistent world map yet, so closing and reopening the guide means re-tapping the alignment corners.
• Ghosts are always axis-aligned boxes. Cylinders, bags, and irregular items render as their bounding box. The shape is preserved on the item — it's just not visualised in the ghost.
• One anchor for the whole plan. If the container physically moves after you align, every ghost is wrong until you re-align.
• No placement validation. When you tap "Placed," the app takes your word for it. It doesn't check that the item is actually inside the ghost's volume.

The solver

• Collision checking uses bounding boxes. Items are fit-checked as their bounding box, so a cylinder or bag that's genuinely smaller than its box doesn't get credit for that space. Shape Type still affects orientation and access decisions.
• It's a heuristic, not a proof. Easiest Load, Tightest Fit, and Best Access are three good strategies, not optimal ones. For a very tight pack, try Tightest Fit. Unusual input can still produce a plan you'd improve by hand.
• No weight balancing. Heavy items end up low because the stackability rules force it, but the solver doesn't try to centre mass or distribute weight across the space.
• Load-bearing is binary. If you mark an item "supports weight," anything can stack on it. Reality has gradients; the solver doesn't model them.

Find a Container

• Catalog only. Recommendations are scored against a fixed set of common containers — boxes, totes, suitcases, shipping boxes, and vehicle spaces. If your real container isn't in the catalog, you won't see it. Measure it once with Start new pack instead, and with Pro it's saved for reuse.
• Single container only. It won't propose splitting your items across two boxes. If nothing in the catalog fits, the answer is "no good fit" rather than a multi-container plan.

Storage & sync

• Local, with two opt-in exceptions. Saved packs, spaces, and items live on the device that created them. There's no iCloud sync in this release, so a delete-and-reinstall or a new device starts empty. Pro users can export to files first as a backup. Two features do reach a network: PackPilot fetches the Community Contributed catalog from a server, and if you opt in to contribute, submitting a scan sends its dimensions and metadata (never photos, location, or free-typed text) to a moderation server.
• No history. A saved pack is its state at last save. There's no record of "I packed this trunk on 3 June."

Purchases

• Two ways to get Pro. The 30-Day Project Pass is a non-renewing subscription that supplies 30 days of Pro — it ends automatically and is not charged again (never "per month," never auto-renewing). Pro Lifetime is a one-time non-consumable purchase for permanent access. Live Scan is an entitlement included with either; there is no separate Live Scan product. Final launch pricing is not yet live — the App Store shows the authoritative localized price at checkout.
• Restore is best-effort. If Apple hasn't propagated your purchase yet, Restore reports "No prior purchase found" — the app can't distinguish that from having genuinely never bought it. Wait a moment and try again. Pro Lifetime restores permanently; an active Project Pass restores only while its 30-day access period is still valid.
• Free-tier caps are enforced on the device. Clearing and reinstalling resets the saved-pack count and does not create another complimentary Live Pack. There is no cloud sync to defeat, and we'd rather not build an account system just to police it.

Platform & accessibility

• English (US) only. No localisation infrastructure is wired in yet.
• No formal accessibility audit yet. SwiftUI gives the controls sensible VoiceOver labels, but the flows haven't been hand-tested with VoiceOver, the largest Dynamic Type sizes, or Reduce Motion. This is a priority before 1.0 ships.
• Some copy is still being refined. A few onboarding and tooltip strings read awkwardly. If something confuses you, tell us — that's the fastest way it gets fixed.
• No third-party analytics or advertising SDKs. Which means we genuinely don't know what you're struggling with unless you write to us. Please do.

---