The Code Lens

The Visual Thinking Lens, turned on source code as a structural field. Code carries mass — brackets and keywords are heavy, whitespace empty. We weigh every structural dimension as a line, learn the norm from a corpus, and read the presses — where this code departs from the middle. A coherent departure on several lines at once is improbable by luck: a deliberate structure (potential energy). Incoherent structure is the junk gate's tangled. Quick to default; the rare corner is rarely reached.

Source — paste a function, a class, a config Language

Learn the norm from your own corpus

Paste real code samples separated by a line of ---. The lens refits its field — so a press comes to mean unlike what you write, and the default means you're repeating yourself.

Paste code (or pick a preset) and run the lens.

The mass field — where structure concentrates

Each character is weighted by its structural role. The cross is the mass centroid; the line is its drift from the geometric center; the ring is the radius holding half the mass. Hover the field to inspect any character's class and mass.

Reading the position

Run the lens to read how this code sits relative to the learned topology.

Deterministic. No upload, no network, no parse of meaning — only structure. Eight structural lines (line length, indent depth & variance, vocab richness, structure density, literal·comment ratio, nesting depth, control density) are robust-z'd against a field learned from a corpus of ten common patterns — or from your own code, so a press means "unlike what you write." choice = Σ relu(|z|−1.5): one hard press (|z|≥2.5) ≈ three semis. internal_coherence is the independent junk gate — balanced brackets, consistent indentation, well-formed identifiers; unbalanced structure can only be tangled, never potential energy. Tokenizing follows the language pack (JS/TS, Python, Rust, Go, SQL). Lineage: VTL mass-map · the interestingness choice/press model · the junk gate.