@yonyoniz tests Jev on Hebrew tweets: same stance as English on 8 of 9; language isn't the bottleneck.

Benchmark-style X article with cover image. Nine of the author’s Hebrew tweets (plus English translations) were scored on jev-1.13.0 for stance and sarcasm; 8 of 9 matched stance across languages (~0.3s/call, $0.0005 for 18 calls). Idioms can shift sarcasm; low confidence flags mismatches. Author calls it a probe, not a full benchmark. Playground: https://play.yonyon.ai/jev-hebrew-2026-10-24/