# Co-Design Bench shows AI coding agents fully finish only about one in four real tasks

> Original title: AI coding agents still need human cleanup on nearly three out of four tasks in a real-world test

- Source: The Decoder
- Published: 2026-11-08T09:50:59.000Z
- Bad Influence: https://badinfluence.news/items/xmlox7912n6k0iwkd2h69accs
- Original: https://the-decoder.com/ai-coding-agents-still-need-human-cleanup-on-nearly-three-out-of-four-tasks-in-a-real-world-test/

## Summary

Researchers from Alibaba and several Asian universities built Co-Design Bench, a benchmark of real user change requests for coding agents. GPT-6 Astra, the best of ten models tested, meets about 70 percent of task requirements on average but is fully correct in only 26.4 percent of tasks. The authors say performance drops as tasks combine more types of changes, and that results also depend on the harness the model runs in.