Rosetta M

One computation, three faces, and time. Each stone is one small machine-learning computation shown as its Math, the proposed M code, and its Visual (structure and live data), with time as the fourth dimension; some stones add a Trace face, the time axis itself. Choose a stone above; drag it to rotate, or lay its faces side by side; scrub the steps and the numbers move while the equation, the code and the dataflow stay put; walk the equation symbol by symbol and watch each one light up on every face. Rest the pointer on anything for an explanation; press ? for the keys.

M is proposed, not run. The M face is notation rendered from the same record as the Math face; nothing here parses or evaluates it, and its glyphs are placeholders. Every number on the page comes from a real training run in sw-MLPL, replayed from a recorded trace.

Math

M proposed notation · placeholder glyphs

Visual structure and data at this epoch

Trace the time axis

    Rotate
    View 100%
    Travel
    Walk

    Focus

    Click any symbol on any face, or in the readout, to focus it everywhere. Esc clears.

    Keys