Research
An AI lab for the
human mind.
TetiAI is an applied AI research lab. We don't chase benchmarks — we work on a single question: how can an AI remember everything about you, help you every day, and leave your mind stronger than it found it?
Our north star
Our long-term goal, by 2030, is the first Symbiotic AGI: a form of general intelligence that extends the human mind rather than replacing it.
That framing changes what we research. Instead of asking "how do we make the model smarter?", we ask "how do we make the human-AI pair smarter?" — which leads to very different problems: memory that persists for a lifetime, systems that detect cognitive over-reliance, and interfaces that strengthen reasoning instead of shortcutting it.
What we work on
Persistent memory
Most AI systems forget everything between sessions. We research lifelong memory architectures — memory graphs, retrieval, consolidation — so an assistant can build a continuous, private understanding of one person over years, not minutes.
Cognitive health
Heavy AI use can quietly erode reasoning, memory and original thought. We study how to measure that risk in real interactions — engagement, autonomy, verification, long-term drift — and how an AI should adapt its behavior to prevent it.
Human-AI symbiosis
We design interaction patterns where the AI scaffolds thinking instead of replacing it: adaptive guidance levels, Socratic dialogue, suggested pauses. The goal is an assistant that leaves you sharper after every session.
Open-model adaptation
The Teti series of models is built on open-source foundations. We work on post-training and system-level adaptation of open models — so our stack stays inspectable by independent researchers, and personalization never requires training on user data.
How we work
Applied, not abstract
Everything we research ships in Teti. The product is the lab bench: real users, real long-term memory, real cognitive signals — under strict privacy guarantees.
Open by default
Lucid, our cognitive-health layer, is open source. Our system prompt is public. Our Charter is public. If a result matters, we'd rather publish it than hide it.
Small and specialized
We are a small team going deep on a narrow set of problems, rather than a big team going wide. Depth over headcount.
Bound by the Charter
Our research is constrained by our Ethical Charter and overseen by an independent Ethics Committee with veto power. No military, no surveillance — including in research.
In the open
Lucid
Our open-source cognitive-health layer: real-time adaptation, fatigue detection and the Cognitive Drift Index.
ExploreSystem Prompt
We publish the instructions that shape Teti's behavior, so anyone can verify how it's meant to act.
ReadResearch notes
Updates, findings and release notes from the lab, published as we go.
ReadInterested in what we're building?
Try the product our research ships in, or get in touch if you want to collaborate.