Ernst AI / Labs
A research practice for building AI that can reason, remain accountable, and help people act with greater clarity.
03
frontier inquiries
01
shared alignment practice
∞
questions worth asking
Active inquiries
Labs is where principles become prototypes. Each inquiry is designed to produce a useful method, not just a provocative idea.
Truth-bound systems
Researching Ibn Arabi’s Oneness of Being as a conceptual basis for recursive language models that preserve context, relationship, and truth across complex decisions.
Output
A testable framework for relational reasoning and truth-bound evaluation.
Plural perspectives
A simulation space inspired by the ancient Egyptian Day of Judgement: perspectives are reconstructed as agents and asked to account for their values, claims, and consequences.
Output
A participatory method for examining judgment, governance, and accountability.
Human state
Investigating how subjective state, language, and decision quality might be represented responsibly—without turning spiritual or psychological ideas into unsupported measurement claims.
Output
Careful interaction patterns for reflective products and solutions.
Research position
We treat AI alignment as the connective tissue between research, solutions, and products: a disciplined way to translate values into evaluation, interaction, and governance.
That means reinforcement learning, truth-bound reasoning, and human oversight belong in both our internal experiments and our client work. Labs tests the ideas; Solutions makes them usable; Products puts them in people’s hands.
See the operating modelA living program
Our journal documents what we learn as these questions move from philosophy to method to working systems.
Read the journal