BrickStack / Inventory experiment
I built the LEGO inventory tool I could not find.
A personal collection problem became a store-ready product for identifying, valuing, locating and selling LEGO parts.
Live productVisit BrickStack
The reason
The collection had outgrown my memory.
LEGO was spread across bins, drawers and bags. I no longer knew what I owned, where it lived or what it was worth. I bought duplicate parts and left valuable stock sitting unidentified.
I also sell excess LEGO through BrickLink. Its inventory workflow is powerful, but listing a part meant finding it, setting condition, location, quantity and price through a click-heavy process. The friction meant the pile kept growing.
The experiment
Could one product person ship the whole thing with AI?
The second motivation was professional curiosity. I wanted to test how far one person could take a meaningful product with AI accelerating the work.
This could not be a disposable prototype. It needed camera access, external APIs, authentication, payments, store approval and enough operational reliability for me to use it on my own inventory.
The product
Make the fiddly workflow feel obvious.
BrickStack turns the workflow into a short loop: identify a brick, confirm the details, see its value, record its physical location and synchronise the listing.
- 01Identify parts from the camera without relying on tiny mould numbers.
- 02Search parts and sets through the Rebrickable API.
- 03Lock repeated fields in batch mode to process a whole drawer quickly.
- 04Track condition, storage location, personal collection and store stock.
- 05Synchronise listings and inventory directly with BrickLink.
The hard part
Building was the easy 80%. Shipping was the other 80%.
I began in SwiftUI, then moved to Next.js and Capacitor so web, iOS and Android could share one codebase. That decision improved long-term leverage, but it created a difficult native integration layer.
Camera access, Firebase messaging and StoreKit did not simply work through the wrapper. Build-time patch scripts kept the iOS project stitched together, while Apple's agreements, tax forms, receipt validation and review rules added a second product workflow behind the product itself.
The outcome
The experiment became a product I could actually use.
BrickStack shipped on the web and iOS with the complete inventory loop in one product. The original pile of unidentified stock became a practical test bed for scanning, pricing, storage and marketplace integration.
More importantly, the project made the limits of AI-assisted development visible. Generating working pieces became faster; integrating them into a reliable store-ready product still demanded product judgement and persistent last-mile problem solving.
What it taught me
AI shortened the distance between wishing and shipping.
AI lowered the barrier enough for ‘I wish this existed’ to become ‘so I built it.’ It did not remove the human work of integration, judgement and launch. The last mile is still where most of the value lives.