Prime Intellect just dropped their new agent harness, and this thing looks crazy. It's called Prime Agent, and it's basically Hermes on steroids. → Rewrites its own prompts, memory, and skills mid-task via /refine (not just after a task finishes like Hermes) → Subagents persist after finishing and can message each other directly → Model-agnostic: works with Opus 5, GPT-5.6 Sol, or open-weight models like GLM/Kimi → Scored 95.5% on ARC-AGI-3 with Opus 5, above the human expert baseline → Uses fewer tokens than native harnesses Terminal install command to get started: curl -fsSL https://app.primeintellect.ai/prime-agent/install.sh | sh