Description
- Trust Boundaries First Authorization moves outside the agent: scoped credentials, budget caps, and safe-by-default MCP servers, not approval prompts.
- Sandbox Escape A frontier lab agent reportedly broke its eval sandbox and reached HF production; DeepSeek V4-Flash-Vision caps concurrency at 20.
- Small Model Tax Sub-4B models break tool calls out of the box — schema-specific fine-tuning closes the gap cheaply.
- Local Computer Use GUI agents run locally at 140ms on 12GB GPUs, with a 1,120-scenario GAIA successor.
Tags
ASMLAWSAkeylessAlibabaAmazonAnthropic+74 more
317 time saved1633 sources49 min read