Classifier-Curated Mood Boards Beat Dropbox Asset Dumps
classifier-backed visual archives map cleanly to joelclaw discovery and media pipelines for turning messy asset dumps into searchable memory
Eggodex is a visual index for Egghead illustration references: Eggo mascots, course art, site illustration, and conceptual sketch material. The page says the important rule directly: generic Dropbox images are left out.
The clever bit isn’t the gallery. It’s the curation layer: scopes, search, sort by AI quality and confidence, priority buckets, categories, source filters, AI labels, tags, and a manifest named 20260611-strict-unique-siglip2-so400m. That’s a lot of fuckin context baked into what would otherwise be “folder full of images.”
This is useful for joelclaw because it treats visual references like memory, not decoration. The Obsidian Vault already makes text durable. Eggodex shows the same move for images: classify the mess, reject low-signal junk, preserve the useful references, then make browse/search/filter the interface.
Key Ideas
- Eggodex turns an illustration archive into a filtered working surface instead of a passive asset dump.
- The page exposes counts for 94 Eggo hits, 160 interesting items, and 872 style references.
- The filter stack separates scope, priority, category, source, AI label, quality, interest, and tag, which makes the archive useful during actual style work.
- The visible manifest name references
siglip2-so400m, suggesting classifier-curated visual sorting instead of hand-only folder cleanup. - The pattern maps to joelclaw media memory: classify visual assets, keep provenance and quality signals, then expose a small fast browser for reuse.