News Challenge Category Strategy

Target

50 published news fact challenges per day, 1 challenge per fact.

Provider

Event Registry (NewsAPI.ai) — full article body on free tier (avg 3,000-5,000 chars per article).

Token Budget

MetricValue
Target facts/day50 publishable
Articles needed/day~84 (accounting for quality + notability yield)
Search tokens/day~5 (20 articles per token)
Tokens/month~150
Free tier allocation2,000 (one-time)
Runway at free tier~13 months

Starting Categories (5)

#Search KeywordEko TaxonomyAvg Body LengthRationale
1sciencescience5,093 charsHigh notability, verifiable facts (studies, papers, discoveries)
2technologytechnology4,436 charsConstant news flow, concrete facts (product launches, funding, specs)
3sportssports5,167 charsHighest content quality, extremely verifiable facts (scores, dates, records)
4healthhealth-medicine4,778 charsStrong reader interest, citable sources (journals, WHO, CDC)
5businessbusiness4,066 charsQuantitative facts (revenue, deals, market data), solid notability signals

Pipeline Yield (observed from smoke test)

StageYield
Articles fetched → pass 400-char quality filter~90%
Usable articles → facts extracted~100%
Facts → above 0.6 notability threshold~67%
Facts → pass validation~100%
End-to-end: articles → publishable facts~60%

Categories Deferred

KeywordEko TaxonomyReason to defer
entertainmententertainmentLowest content quality (3,175 chars avg), facts tend toward opinion/gossip
politicsgovernmentsHigher risk of bias-related validation issues, subjective facts
current-eventscurrent-eventsToo broad as keyword; other 5 categories already capture newsworthy events

Scaling

Adding categories requires no code changes — update active topic categories in the DB and the cron dispatches automatically. Each additional category adds ~1 token/day (~30 tokens/month).

CategoriesTokens/monthFree tier runway
515013 months
82408 months
103006 months

The $90/month 5K plan becomes worthwhile when running 10+ categories at higher frequency, or when historical data access (back to 2014) is needed for evergreen fact enrichment.