KOHER

AI handles language.
Code handles judgement.

Small tools that help you see your own work clearly. Free for students. Open source.

AI qualifies · Code judges · AI narrates

Try the tools Support the practice

Co-created by Prayas Abhinav + Claude Code

Language is probabilistic, fluid, generative — AI handles it well. Judgement is deterministic, rule-bound, auditable — code handles it well. They require different treatments. Asking a language model to make judgements is asking a probabilistic text generator to produce deterministic verdicts. The work may come out right, but there is no way to know when, and no way to adjust what you cannot inspect.

The difference in practice

Design concept evaluation
Typical AI approach
Black-box verdict. No visibility into criteria. No way to learn or adjust.
Koher approach
AI qualifies input → Code applies explicit thresholds → AI narrates the result. Each layer is auditable.

The separated version is auditable, consistent, and adjustable. When standards change, you update the rules — not retrain the AI.

Small things that might help.

Each tool does one thing. Free to use. Source available. See all tools →

Writing & Thinking · Mar 2026 Live

Fragment Mapper

Paste scattered text fragments. See where they cluster, diverge, contradict, or echo. Four signals measure meaning, vocabulary, emotion, and voice. Six deterministic rules surface structural relationships. Claude Haiku narrates what the rules found.

Try it → GitHub
Play & Games · Feb 2026 Live

Play Shape Diagnostic

Select three experiential qualities from twelve — anticipation, tension, relief, discovery, and more. See whether your combination is harmonic, distinct, dynamic, complex, or paradoxical. User selection replaces AI classification; embedding similarities determine relationships.

Try it → GitHub
Design Concepts · Feb 2026 Live

Coherence Diagnostic

Paste a design concept. Get a diagnostic across five dimensions — what's solid, what's thin, what's unclear. A domain-trained DeBERTa model qualifies the text, deterministic rules apply judgement thresholds, Claude Haiku narrates the result.

Try it → GitHub

Open source. Coherence, Play Shape, and Fragment Mapper are MIT licensed. Tools published after 5 April 2026 are AGPL-3.0. Verify your email to try hosted demos (10 free analyses, no account). Or clone any repo and run locally.