Get real-time correction while writing in a language you're learning
Fix the actual mistakes without your sentence getting rewritten into something a native speaker would never have said.
- Use it for
- Language learners writing or messaging in a language they're still learning.
Asking a model to "fix my French" usually gets you a fluent native speaker's version of what you were trying to say — which is useful for reading, but useless for learning, because it replaces your actual mistakes with a fluent rewrite you didn't produce and won't remember. The thing that helps you improve is different: your sentence, corrected at the exact points where it was wrong, with the wrongness explained, so the next time you write that construction you get it right without help.
This prompt asks for that instead — minimal correction plus explanation, not a rewrite.
When not to use this
If you actually need a polished, correct message to send right now — a professional email in a language you don't speak well — a full fluent rewrite is the right tool, not this. This is specifically for practice, where the point is what you learn from the correction, not the output itself.
Did this work?
You understand why each correction was made, not just what the corrected version says — and your original sentence's meaning and length survive the correction intact.
Tested on claude-opus-5. Evidence status is draft; it moves to battle-tested only on recorded runs, never by hand.