Claude consistently drops Spanish accents/tildes despite explicit instructions

Description

Claude Code consistently omits accents (tildes) when writing in Spanish, producing text like "investigacion", "priorizacion", "critica" instead of "investigación", "priorización", "crítica". This happens across all Spanish output - social media posts, comments, messages.

Reproduction

1. Set up memory/instructions explicitly requiring proper Spanish accents
2. Ask Claude to write anything in Spanish (e.g. post comments on Spanish-language platforms)
3. Output consistently lacks accents on words that require them

Expected behavior

Spanish text should include proper orthographic accents, the same way English text includes correct spelling. Words like "también", "información", "está", "más" should never appear as "tambien", "informacion", "esta", "mas".

Actual behavior

Claude drops virtually all accents in Spanish output. This persists even with:

• Explicit CLAUDE.md instructions
• Persistent memory files marked as CRITICAL
• Multiple user corrections across conversations

Notes

• This is not an encoding issue - Claude can produce accented characters when specifically asked
• The problem is that the model defaults to unaccented Spanish despite instructions
• It's the equivalent of consistently misspelling common English words
• Model: claude-opus-4-6 (1M context)

🤖 Generated with Claude Code