← Binderdex dev log

2026-07-16

Coverage checker for binder idea gaps

tcgdataproduct

How many binder pages does each card actually have?

I shipped a set-coverage checker for Binderdex this week, and it answered a question I'd been squinting at: when a new set drops, which special illustration rares are under-served by the card-page recommender?

Here's the thing about binder ideas. Each IR or SIR card gets catalog pages built from related cards that share overlap tags. The recommender surfaces those pages. But until now I had no quick way to ask, per card, "how many published pages does this card actually appear on?" Some cards might have six. Some might have zero. The zero cards are the gap list, and that's where I need to generate new subjects.

The checker is read-only and pretty small. It walks every full-art candidate in a set, calls the same published-catalog lookup the site uses, and prints a report:

pnpm --filter content-generation binder:set-coverage -- \
  --set-slug pitch-black --min-ideas 3 --json

The output is a per-card row with the idea count and the overlap tags the recommender found. Cards below --min-ideas are the gaps. Feed their themes into the subject generator, rebuild, swap, done.

What I like about this is it turns a vibe ("I think this set feels thin") into a number per card. The gap list is actionable. And because it's read-only and uses the same lookup path as the live site, what it reports is what a visitor would actually see. No drift between the audit and reality.

The next step is wiring it into the new-set workflow so coverage gets measured right after ingestion, before any card ships without its supporting pages. But even as a standalone script it already changed how I triage a set launch: run the checker, read the gap list, generate subjects for exactly those cards.

this is the build log of binderdex · www.binderdex.com · all entries · essays