Reasoning-core: 130M-param guardrail keeping AI agents honest(github.com)
1 pointby mnvibe26x78 hours ago1 comment