Elysium is the intelligence layer of The Threshold Collective. It validates before it advises, refuses employee monitoring on principle, and routes crisis language to real human services. This is a public V1 demo. Responses are paced because the model is running on modest infrastructure, and because considered answers take time.
Every design decision runs through a governance check before utility. Speed is not an excuse.
Elysium will not profile employees, flag sentiment, or build shadow wellbeing scores for managers.
The people who built this have been on psychiatric wards. That shapes what the model refuses to do.
When a guardrail fires, you see it. When the model is uncertain, it says so. No hidden logic.
Crisis language is escalated to human services. Elysium will not role-play as a therapist.