{"slug": "ai-coding-agents-ace-isolated-tests-but-complete-fewer-than-one-in-ten-tasks-in", "title": "AI coding agents ace isolated tests but complete fewer than one in ten tasks in a simulated workplace", "summary": "A new simulated-workplace benchmark found that AI coding agents, including Opus 5.5 and Astra, completed fewer than one in ten tasks end-to-end when required to work autonomously through messy daily workflows with colleagues, project boards and incomplete handoffs, according to The Decoder. The models finished clear, standalone assignments quickly but lost track when acting as autonomous developers.", "body_md": "A new real-world test puts AI coding agents in a simulated developer job with colleagues, project boards, and incomplete handoffs, where they have to keep developing software on their own and reliably close open tickets in a messy daily workflow. Even Opus 5.5 and Astra fall far short, solving fewer than one in ten tasks completely. The models finish clear, standalone assignments quickly, but they lose track as autonomous developers.\n\nThe article [AI coding agents ace isolated tests but complete fewer than one in ten tasks in a simulated workplace](https://the-decoder.com/ai-coding-agents-ace-isolated-tests-but-complete-fewer-than-one-in-ten-tasks-in-a-simulated-workplace/) appeared first on [The Decoder](https://the-decoder.com).", "url": "https://wpnews.pro/news/ai-coding-agents-ace-isolated-tests-but-complete-fewer-than-one-in-ten-tasks-in", "canonical_source": "https://the-decoder.com/ai-coding-agents-ace-isolated-tests-but-complete-fewer-than-one-in-ten-tasks-in-a-simulated-workplace/", "published_at": "2026-11-22 15:15:48+00:00", "updated_at": "2026-11-22 15:43:30.922013+00:00", "lang": "en", "topics": ["ai-agents", "ai-research", "artificial-intelligence", "large-language-models"], "entities": ["The Decoder", "Opus 5.5", "Astra"], "also_reported_by": [], "takeaway": "Agents that pass isolated coding tests solved fewer than one in ten simulated-workplace tasks, so evaluate them on multi-step ticket closure with incomplete handoffs rather than standalone benchmarks.", "recommendation": {"advice": "Test coding agents on end-to-end ticket closure with incomplete handoffs and project-board context, not isolated tasks; the simulated workplace showed fewer than one in ten tasks completed.", "playbook": "evals-testing", "playbook_url": "https://wpnews.pro/playbooks/evals-testing"}, "alternates": {"html": "https://wpnews.pro/news/ai-coding-agents-ace-isolated-tests-but-complete-fewer-than-one-in-ten-tasks-in", "markdown": "https://wpnews.pro/news/ai-coding-agents-ace-isolated-tests-but-complete-fewer-than-one-in-ten-tasks-in.md", "text": "https://wpnews.pro/news/ai-coding-agents-ace-isolated-tests-but-complete-fewer-than-one-in-ten-tasks-in.txt", "jsonld": "https://wpnews.pro/news/ai-coding-agents-ace-isolated-tests-but-complete-fewer-than-one-in-ten-tasks-in.jsonld"}}