01
Scan
Photograph a shelf of boxes. Vision identifies the sets and adds them to the backlog.
← Work / ProductsA catalog knows sets.
AI-native product · designed, built & operated solo
A catalog knows sets.
The Archivist
knows you.
An AI-native backlog manager for serious LEGO builders. Scan boxes, schedule weekends, track every bag, and ask an Archivist who remembers the collection and the person building it. Solo-built, ~140K lines of TypeScript, live on web and iOS.
// product brief
- ROLE
- Sole architect + developer
- STATUS
- Live on iOS
- MODEL
- Subscription product
- SCOPE
- Mobile → backend → ops
The Brick Archivist
The product character
He remembers the collection better than you do.
The Archivist is not a generic assistant wearing a themed prompt. He knows every set, every tracked bag, every abandoned build, every retirement date, and the builder's actual pace. Ask what fits a free Saturday and the answer comes from evidence.
You · “What should I build this weekend? I have about three hours.”
The Archivist can identify the sets that fit, account for in-progress bags, estimate remaining time from personal pace, and explain the recommendation.
140K+
lines of TypeScript
85+
database migrations
05
product surfaces
iOS
shipped May 2026
01 — The gap
Databases remember sets. None remember the collector.
Serious collectors live across a stack of databases — Brickset, BrickLink, Rebrickable, Brickeconomy. They are catalogs: look something up, leave. None of them know your collection, your build habits, or whether the set you are eyeing is actually worth it for you. The hobby was drowning in reference tools and had no companion. Bricktelligence starts where reference databases stop: it accumulates context, recognizes behavior, and turns that memory into useful action.
02 — Five connected surfaces
02
Schedule
Give it the available weekend hours and get a build calendar spread across real time windows.
03
Track
Follow every bag, note, photo, and session. Pace estimates improve as the system learns the builder.
04
Ask
The Archivist answers with collection history, current progress, retirement dates, and personal context.
05
Share
Finished builds become journal entries, gallery posts, build cards, and a public builder profile.
03 — Intelligence architecture
Memory is the product—not a chatbot bolted onto it.
Signals
01- Vision intake
- Build events
- Collection data
- User questions
Memory
02- Narrative facts
- User aggregates
- Event firehose
- Inference layer
Intelligence
03- Agent tool loop
- Verdict engine
- Recommendation context
- Rate limits + telemetry
Experience
04- Scan
- Schedule
- Track
- Ask
- Share
Events become durable facts and per-user aggregates. The inference layer turns behavior into meaning. The agent loop can query tools directly, while the verdict engine combines memory with structured set data to make a personal buy, wait, or pass call.
04 — What shipped
Product, intelligence, commerce, and operations.
One system across Expo and React Native, Next.js, Supabase, vision intake, agentic AI, Stripe, RevenueCat, and the production telemetry needed to operate it.
- ▹The Archivist — a memory-driven co-pilot on a persistent narrative fact store + per-user aggregates, fed by an event firehose
- ▹Agentic ask-loop with tool access, rate limiting & telemetry
- ▹Inference layer (behavior → facts) + verdict layer returning Worth-it / Wait / Pass on any set
- ▹Vision-based Capture: camera → set ID, number & piece count → collection
- ▹Five surfaces (Capture · Vault · Forge · Discover · Home) across the Know / Build / Discover pillars
- ▹Shipped to iOS May 2026 (Android in progress) · Expo/React Native + Next.js + Supabase (85+ migrations) · Stripe + RevenueCat · ~140K lines TypeScript
Constraints / Evidence
The conditions shaped the system.
- ◆Solo ownership across product design, mobile, web, backend, AI, commerce, and operations
- ◆AI recommendations must be grounded in personal history rather than generic catalog data
- ◆One experience must stay coherent across browser, iOS, subscriptions, and asynchronous inference
- ◆A niche enthusiast product has to earn complexity feature by feature
iOS
shipped product
A consumer product available beyond a portfolio demo.
140K+
TypeScript lines
Product, intelligence, commerce, and operations built solo.
85+
migrations
A data model evolved through sustained product development.
5
connected surfaces
Capture, Vault, Forge, Discover, and Home share one memory system.
The result
A live consumer product where AI earns its place by remembering, inferring, and acting on context over time.
Talk product systems →