During December 2025, Onozaty focused on resolving a UTF-8 multibyte handling bug in the prisma/prisma repository, specifically addressing issues in the byline module when non-ASCII characters spanned stream chunk boundaries. He replaced the previous chunk.toString approach with Node.js’s StringDecoder to correctly buffer incomplete multibyte sequences, ensuring reliable generator output across languages, including emoji and Japanese text. Onozaty contributed changes across multiple packages—generator-helper, client, internals, and migrate—while standardizing built-in imports for compatibility. He reinforced the fix with comprehensive TypeScript-based tests targeting edge cases in stream processing, demonstrating careful attention to correctness and cross-package robustness in the solution.