Reasoning-core: 130M-param guardrail keeping AI agents honest

(github.com) by mnvibe26x7 | May 13, 2026 | 0 comments on HN