Six questions we're building answers to. In cognition, language, context, and the architecture underneath.
The cognitive architecture running underneath Living Mirrors.
Every AI runs the same cognitive architecture. Analytical. Convergent. Sequential. The same mind, for every person, on every problem.
We build the other architectures. Cognitive mode specifications that restructure how intelligence perceives, connects, and reasons. Seven principles from neuroscience. Five emergent combinations.
Plain text as cognitive program. The document changes the shape of thought.
An AI that helps you think is one thing. An AI that helps you choose is another thing entirely. Seven principles grounded in Kahneman, Klein, and Bayesian cognition.
When the mirror knows your decision architecture better than you do — should it tell you? If it frames options, the framing shapes the choice. Surfacing is not neutral.
The deficit was always the surplus.
The meaning engine
Vervaeke diagnosed the meaning crisis: we lost the frameworks that provided meaning and replaced them with nothing. Quest is a direct response.
A Virtue Compass measures not how far you walked but how you walked. Panksepp's SEEKING system fires during pursuit, not arrival. Quest focuses on departure.
The operating system is open. The marketplace is where it compounds.
Every conversation you have with AI starts from zero. Your decisions, your history, your knowledge — gone. You rebuild the same context every single time.
A walnut is the fix. Five plain-text files that carry identity, state, history, knowledge, and work queue between sessions. Any AI loads them. Every session compounds. The context never leaves your machine.
Walnut World is the ecosystem. Skills, cognitive modes, and infrastructure that anyone can build, share, and remix. Open architecture. No lock-in.
Explore Walnut WorldThe sovereign context layer. File-based. Runs on your machine. Nothing phones home. Nothing leaves without your say.
Ben Flint built the framework — memory, sessions, save protocol, the caretaker runtime that makes any AI agent interchangeable. Attila Mora built the perception layer — nine cognitive modes across five axes that restructure how intelligence sees, connects, and reasons.
The research improves the tools that make the research possible. Two minds. One architecture. Open source forever.
What if the language you write a cognitive specification in changes how the AI thinks?
Agglutinative languages encode relationship, temporality, and intentionality within single morphological units. One word contains what English needs an entire clause to express. Hungarian. Estonian. Turkish. Japanese.
Neural networks build meaning through layer composition. These are structurally homologous. The morphological layering of an agglutinative word mirrors the layer composition of a neural network. The language of the document is the programming language.
The first experiment: write the Savant Mode specification natively in Hungarian. Run identical prompts through both. Measure the outputs. Academic partner validates. Results published open source.
Be the first to see the research proposal when it opens.