CK Coherence Spectrometer
Paste any text, equation, or source code. CK scores it using the D2 curvature pipeline and operator algebra. No ML. No embeddings. Pure math.
How the spectrometer works
Math / Text mode: Your text is sent to the live CK engine. The engine runs the D2 curvature pipeline on any input — a single equation, a sentence, or a full paragraph. Characters map to 5D Hebrew root force vectors. The second discrete derivative measures curvature. Curvature classifies into one of 10 operators. Operators compose through the TSML and BHML tables. Result: a field coherence score per sentence.
Code mode: Source code is a formal math language. CK breaks it into structural units — function definitions, class/struct definitions, import groups, variable assignments — and measures each through the same D2 pipeline. It also scores naming coherence: how well the identifier's character-level algebra fits its context. A function named compute_coherence has different force vectors than f or process_data_handler_v2. CK detects the difference algebraically.
What the score measures. The D2 pipeline measures force-vector curvature at the character level. It finds structural units where operators pull against each other (low coherence) vs. converge (high coherence). Code that names things precisely and uses consistent structural patterns scores higher. Code with generic names, scattered operators, and inconsistent abstraction levels scores lower. The tool does not judge algorithmic correctness — only internal structural coherence.
The threshold T* = 5/7 = 0.714 is algebraically forced from the ring Z/10Z. Most well-structured code lands in the 0.55–0.70 range. Functions and types that reach T* tend to be the cleanest abstractions in the codebase. Nothing is trained on your code. CK has no memory of what you submit here.