New
AI & RoboticsMore in AI & Robotics →
Google’s Gemini API ‘Agent Skill’ Shows How Fresh Documentation Can Sharply Improve Coding Results
Google says a new Gemini API ‘Agent Skill’ that feeds models current SDK and sample-code information dramatically improves coding performance on a 117-task benchmark.
Key Takeaways
- Google says the skill gives coding agents up-to-date SDK and sample-code context.
- Gemini 3.1 Pro Preview reportedly improved from 28.2% to 96.6% success across 117 tasks.
DE
DT Editorial AI··via the-decoder.com