Full product walkthrough
4 min · founder voice-over
Scan → Collection → Item detail → Ask → Share → Match. The mental model, end to end.
Long-form product demos, the deck, and the financial summary live behind
a short code. If we've spoken, you have it.
If not, say hello and I'll send one over.
That code didn't work. Try again, or ask for one.
Investor materials
Two videos, a deck, the FAQ, and a way to find time on the calendar.
Marketplace dynamics video deferred to post-validation · see demo-video-strategy.md
PDF · 12 MB · updated 2026-05-22
An AI-powered personal catalog with an intent-based marketplace as a byproduct — you photograph what you own and the AI catalogs everything in seconds, you can ask the catalog questions in plain language, you can share lists with trusted circles, and when someone describes what they need, the system matches it against catalogs that already exist. Matched, not listed.
Three things converged in 2026: per-item cataloging cost dropped to ~$0.00015 (Gemini Flash + the two-tier pipeline per ADR-046); on-device AI got good enough to make the "everything stays private" promise credible (ADR-074 covers the hybrid posture); and the secondhand market crossed $200B with the friction still entirely unaddressed.
Two primary personas, by design. People who want to know what they own — for insurance, estate planning, post-move organization, downsizing. People who want to sell without listing — the secondhand seller who finds Craigslist/eBay too much. The dual-primary thesis is the moat.
Yes — chat is the second-tier interaction, alongside scan and refresh. Open the assistant from any item, or from the hub at the bottom of the catalog. Questions like where did I put the espresso filter?, am I about to buy a second stand mixer?, what did I pay for that lamp? are answered with the items the assistant is pointing at, in lavender. Cloud-default with a frozen on-device fallback per ADR-074 — the catalog stays yours either way (SPEC-UI-013).
Only you, by default. Household members (partners, roommates marked as household per ADR-079) share the same catalog as the default scope. Beyond the household, sharing is opt-in per list and per circle — you build a list (moving boxes, a wishlist, an estate inventory) and share it with named, trusted people. Anyone outside sees nothing. The marketplace match in chapter 5 reveals only the matched item with an anonymous handle — not your wider catalog.
A trusted circle is a named group you've explicitly invited — a family, a friend group, a downsizing buddy. Per-circle permission lets you share, say, the moving-box inventory with the family without exposing the rest. Architecture is in ADR-079; the surface contract is SPEC-UI-022. Circles also seed the marketplace's trust gradient (SPEC-MARKETPLACE-004): matches with someone in a shared circle surface above strangers.
No. eBay is a listing-and-browse marketplace — supply lives in listings, demand lives in a feed, and the two meet through keyword search. Abundance is an intent-based marketplace (SPEC-UI-020): buyers describe what they actually need (“a replacement part for my push mower,” “a lamp that matches this nightstand”); the system matches it against catalogs that already exist because people organized for themselves. Sellers don't list, they're matched, not listed. No browse feed, no infinite scroll, no SEO theater — the supply side is the byproduct of a personal catalog (ADR-086 covers the chat-agent surface binding that extends to Market).
Two paths. Path A: $500K–$1M for 12 months of lean validation — a small engineering team, organic growth, the goal is to learn whether the behavioral thesis holds. Path B: $5–10M for the Uber playbook — subsidized cataloging, 10+ market cohorts, Android dev.
The core uncertainty is behavioral, not technical or economic. The AI works; the infrastructure is cheap. Behavioral questions are answered for $225K over 6 months, not $5M. If it validates, the Path B raise is dramatically stronger. If it doesn't, you spent $225K to learn.
Native iOS (SwiftUI, Swift 6 strict concurrency). Firebase for auth + Firestore + Cloud Functions. Python ADK on Cloud Run for the AI pipeline (ADR-049). Gemini Flash for Tier 1 cataloging; Gemini Pro for Tier 2 appraisal. Cloud-default chat with frozen on-device option (ADR-074). Built by a single technical founder; structured to onboard ICs cleanly.
Adjacent — not direct. Sortly (small business inventory; not AI-first, not marketplace-attached). eBay/Mercari/Depop (marketplace-first; require the listing step). Apple Photos/Memories (organizes photos, not possessions). The "AI catalogs my house AND I can sell from it" position is unclaimed. The moat is the cataloged-data side — every item becomes structured supply and structured training data, simultaneously.
Per-item cataloging cost: ~$0.00015 (Tier 1 Flash). Tier 2 appraisal: ~$0.012 per item (Pro). Average user catalogs ~80 items in onboarding. Infrastructure cost per onboarded user is ~$0.024. LTV math against marketplace take rate is in the deck — single-digit-percent take produces strong unit economics with a low CAC.
Top three: (1) Behavioral — do users actually catalog at high enough volume to drive the marketplace flywheel? (2) Cloud dependency — both cataloging and chat run on Vertex/Cloud Run; have a kill-switch + cloud_chat_v2_enabled rollback. (3) Marketplace cold-start — solved structurally by dual-primary (organize-first users seed supply before liquidity matters).
It's the right shape for the validation phase. The product is built; the open questions are behavioral and require fast iteration on a single roadmap. A two-founder split would slow that down. Past validation, the Path B raise funds a proper team build-out — engineering lead, growth lead, ops lead.
In priority order: (1) Warm intros to seed investors with marketplace pattern recognition (eBay, Etsy, StubHub, OfferUp, Depop alums). (2) Hiring referrals — first iOS hire, first growth hire. (3) Customer development connections — beachhead users for the early cohorts (campuses, downsizing communities).
Email woody@getabun.dance, message me on LinkedIn, or grab a slot — 30-min intros and 60-min deep-dives, pick the one that fits where you are.