Under the Hood
Try it freeHow did we achieve such amazing results?
For the tech nerd and the curious mind — here's what powers every listing.
Advanced AI models, layered together
Claude Opus 5
Primary listing brain — reads your photos and writes the title, description, item specifics, condition and pricing rationale.
Claude Sonnet 4.6
High-volume bulk work — bucketing thousands of eBay leaf categories into the right niche prompt.
Gemini 3 Flash (web-grounded)
Live reference research — collector clubs, catalogs, societies and identification guides for every niche.
Grok 4.3 Vision
Independent fallback vision model so listings keep flowing if a primary model is unavailable.
On-device TensorFlow (ml5) + QR/OCR
In-browser image classification, marker-card QR detection and text reading — fast, private, zero upload lag.
Prompts, iterations & testing
A library of niche prompts
Every collectible niche — stamps, postcards, comics, advertising, breweriana and many more — gets its own expert prompt mapped to exact eBay leaf categories.
AI authoring AI
Stronger models draft and refine each niche prompt from researched reference sites, then a universal prompt layers shared rules on top.
Many iterations of testing
Prompts were run against real listings over and over, comparing AI output to what sellers actually published, and tuned until they aligned.
A learning loop
Seller edits are analyzed for recurring patterns and turned into reviewed prompt improvements — so accuracy keeps climbing.