In a significant development, AI models have rapidly advanced from being professionally unreliable to real-world-ready within the past year. An experiment by Mercor demonstrated that Claude Opus 5 flawlessly completed 20 accounting tasks, while human junior accountants achieved an average score of only 37.5%. This marks a dramatic improvement from mid-2025, when similar models scored around 50%. The agentive capabilities of models like Opus 5.5 are now so advanced that they can often complete complex tasks in a single attempt, raising concerns about their profound impact. AI
IMPACT AI models are rapidly becoming capable of performing complex professional tasks, potentially displacing human workers in fields like accounting.
RANK_REASON The cluster describes a research experiment demonstrating a significant leap in AI model capabilities, specifically in professional tasks. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →