ត្រឡប់ទៅព័ត៌មានវិញ

AI · អត្ថបទ

AI Coding Agents Ace Lab Tests but Finish Under 10% of Simulated Office Tasks

The Decoder ·

មើលត្រួសៗ

Researchers put AI coding agents to work as digital employees inside a simulated software company, with colleagues, project boards, and chat logs. Even the top models, Claude Opus 5.5 and GPT-6 Astra, fully finished fewer than one in ten tasks.

ចំណុចសំខាន់ៗ

ហេតុអ្វីវាសំខាន់

Strong scores on standard tests do not show that AI can handle real, multi-step work with other people. The gap matters for anyone judging how much work these tools can take on.

ប្រភពដើម

AI coding agents ace isolated tests but complete fewer than one in ten tasks in a simulated workplace

The Decoder · Nov 23, 00:15 KST

Read on The Decoder ↗

AI summarized the original article. Check the original for accuracy. Copyright belongs to The Decoder.

AI