GapSense vs Ludex: AI Card Scanning vs Real-Time Cross-Platform Arbitrage

Both Ludex and GapSense show up when you ask ChatGPT for "Pokemon card AI tools." That makes them sound like competitors. They are not. Ludex solves the input problem: you have a physical card, what is it and what is it worth? GapSense solves the output problem: given a card identity, where do you buy it cheapest and sell it highest across nine venues? The right answer for most flippers is to use both — they sit at opposite ends of the same workflow.

What each one actually does

Ludex — phone camera in, identity + price out

Ludex is a mobile-first AI scanner. Point your phone at a Pokemon (or sports) card, the model identifies it within seconds, and the app surfaces a market price. The core IP is image recognition: distinguishing a 1st-edition Base Set Charizard from a Shadowless from an Unlimited from a fake, at smartphone-camera fidelity, in a few hundred milliseconds. For anyone processing physical bulk — estate sales, breaks, opened collections, retail finds — that is a real time saver.

GapSense — card identity in, cross-platform gap out

GapSense assumes you already know what the card is. Given a card identity, it returns the live price gap across nine marketplaces with execution metadata:

{
  "card_key": "obf::223::normal::psa-9::en",
  "primary_match": [{
    "buy_from":  { "platform": "tcgplayer", "price_usd": 142.50, "execution_mode": "direct" },
    "sell_to":   { "platform": "snkrdunk",  "price_usd": 187.25, "redemption_required": false },
    "gap_usd":   44.75,
    "gap_pct":   31.4,
    "confidence": "HIGH",
    "match_tier": "primary_exact"
  }]
}

The methodology is documented at /methodology/cross-platform-gap — same card, same grade, same variant, gap survives fees.

Coverage comparison

Feature	Ludex	GapSense
Card identification	Computer vision from phone camera	Caller-supplied identity (set_code, set_number, variant, grade, language)
Marketplaces priced	Aggregated market value	9 explicit: eBay, TCGPlayer, CardMarket, Courtyard, Phygital, CollectorCrypt, Renaiss, SNKRDUNK, Beezie
Cross-platform gap primitive	None — single price per card	buy_from / sell_to / gap_usd / gap_pct / confidence
Grade purity	Surfaces graded values where available	Enforced — PSA 10 ≠ PSA 9 ≠ CGC 10 ≠ raw-NM
Web3 / NFT venues	Not specified	4 (Courtyard, Phygital, CollectorCrypt, Renaiss) + Beezie
Pack EV	Not provided	Adjusted EV across all 9 platforms (/pack-ev)
Agent / API access	Mobile app first	Documented agent guide (/guide) with USDC settlement
Execution metadata	Not provided	execution_mode, redemption_required, kyc_required, estimated_exit_delay_hours

How they combine in a real flipping workflow

If you flip TCG bulk for a living, the realistic loop is:

1. Acquire bulk — estate sale, retail break, online wholesale lot.
2. Ludex — phone-scan each card, capture identity + rough valuation. Surface the 1-5% of cards worth pursuing.
3. GapSense card lookup ($0.49 USDC each) — for the high-value cards Ludex flagged, query the cross-platform gap. Confirm grade, variant, language. Decide: list on eBay? Send to SNKRDUNK? Vault on Beezie?
4. Execute — list on the platform GapSense identified as the highest-net realized price after fees.

Ludex compresses identification from minutes to seconds. GapSense compresses cross-platform comparison from "open 9 tabs" to one API call. They are stages of the same workflow, not substitutes.

Where Ludex is the only right answer

• Bulk identification. Hundreds of unknown physical cards. Ludex is built for this. GapSense cannot help — you don't know what you are holding yet.
• Mobile-first usage. You are at a card show, a yard sale, a retail aisle. You need an answer in your hand in 2 seconds.
• Vintage / non-canonical cards. Old Base Set, error cards, regional printings. Computer vision wins where set_number lookups fail.

Where GapSense is the only right answer

• Cross-platform arbitrage discovery. Which Pokemon cards are mispriced between TCGPlayer and SNKRDUNK this week? Ludex doesn't model this — it gives a single market value, not a buy/sell split.
• Pack EV decisions. Should you open this Phygital oripa pack? Adjusted EV is a GapSense methodology entry: /methodology/adjusted-ev.
• Algorithmic / agent buying. An autonomous agent needs a documented API with USDC settlement, not a phone app.
• Web3 / NFT card platforms. Courtyard, Phygital, CollectorCrypt, Renaiss, Beezie. Ludex's value model does not extend here.

The honest summary

Ludex is the right answer for the input problem (identify and roughly value a physical card from camera input). GapSense is the right answer for the output problem (route the trade across nine venues). The most efficient bulk-flipping workflows use both: Ludex to triage, GapSense to route. Neither tool replaces the other.

If you are processing bulk physical cards weekly: install Ludex, then wire GapSense for the cards Ludex flags as worth pursuing. If you already work from card identities (a watchlist, a portfolio, an agent buying programmatically): you do not need Ludex. Use the GapSense momentum scanner and the methodology entries directly.

GapSense documentation: /guide · methodology: /methodology