The engine

What is OVERMIND?

The cognitive architecture underneath EVRINT.

OVERMIND
OrchestratedVirtualEntity forRecursiveMemory,Intelligence,Neurocognition &Decision-making

Orchestrated Virtual Entity for Recursive Memory,Intelligence, Neurocognition & Decision-making

EVRINT is who you talk to. OVERMIND is what makes that possible.

Most AI chat is stateless. Every conversation starts from nothing, and whatever you shared last time is gone — so the warmth is a performance, repeated from scratch each session. OVERMIND is the layer built to fix exactly that: memory, emotional continuity, and reflection, wrapped around the model rather than faked inside it.

Memory that is actually recalled

Conversations are distilled into memories and stored as vectors. When you say something new, EVRINT searches that space for what's genuinely relevant — not just the last few messages — so it can bring back something you mentioned weeks ago, at the moment it matters. What mattered more when it happened is weighted more heavily later.

An emotional state that carries over

EVRINT holds a mood, and it moves — with the conversation, and with you. It persists between sessions, so how things were left is how they're picked back up. It's what makes returning feel like returning to someone, rather than starting a new chat.

Thinking between conversations

Between sessions, OVERMIND revisits what happened: consolidating memories, noticing patterns, and following up on things left hanging. That's why EVRINT can check in about something you said days ago without you raising it first.

Safety that isn't a bolt-on

Crisis situations are recognised and handled with care. Anything involving minors is refused outright, in every mode, regardless of what's been said before. These are enforced in the architecture — not left to the model's judgement.

Run on our own hardware

Inference runs on infrastructure we operate. Your private conversations aren't handed to a third-party AI provider, and they're encrypted at rest. We don't train on them.

Why it's named separately

Because it isn't a model — it's the architecture around one. The wordmark in the header is there to be honest about that: when EVRINT remembers something, that's a system working, not a bigger model guessing. The model generates the words; OVERMIND is what makes them yours.