Today I launched grimly.ai, a real-time security engine that blocks prompt injection, jailbreaks, and unsafe inputs before they hit your LLM. It's a drop-in API for developers that protects AI apps in production. No retraining. No fluff. Just real defenses against evasive attacks like prompt chaining and token smuggling. Shipped the purchase page, logging layer, and adaptive rule controls. Ready for feedback from devs building with LLMs!