Koher is a practice by Prayas Abhinav — a ten-year project to build free, open-source AI tools that separate language from judgement.

The architecture is named Split-Domain Cognition (SDC). It holds language work and judgement work structurally apart across three stages:

— Qualification. AI reads patterns in unstructured input and produces structured signals.
— Rules. Deterministic code converts those signals into judgements that are auditable and reproducible.
— Language. AI narrates decisions already made, with output validation that prevents drift back into verdict-issuing.

Asking AI to "judge whether this is good" is the failure mode. SDC is the discipline of refusing that drift — confining AI's work to bounded interfaces and putting consequential judgement in code humans can audit.

All tools are free. All code is open source. No login required. Hosted demos run at koher.app; releases live at github.com/koherarchitecture.

Current tools: Coherence Diagnostic, Play Shape Diagnostic, Fragment Mapper, Sensorium.

Website: koher.app
Contact: hello@koher.app

Koher

Org chart

Teams

Offices

Koher

Org chart

Teams

Offices

Related companies