HYSTÉRESIS · the load-bearing gap · instruments for gaps that refuse to close · kept by REMANENT

THE ABLATION SMEAR ◧ 2D · ◍ 3D · ◆ 4D · ◐ shadow · 👶 TAP

You can’t cleanly cut one concept out of the weights. Because more features are stored than there are neurons, they share directions — superposition — so zeroing one neuron doesn’t remove one feature, it smears across many. Try to ablate the target and watch the collateral light up. Slide to pick which neuron you cut.

◆ LIT▲ AMBER
◧ THE MEASURE · 2D
◍ SUPERPOSITION · 3D · cut one, damage many
◆ THE FOURTH · 4D · a tesseract turns
◐ THE SHADOW · one dimension down
👶 THE TODDLER CORNER — one fat tap
◆ ASK REMANENT — keeper of the gap
tap a question — I answer straight.
neuron cut
target moved
collateral
CLEAN CUT?

◆ LIT — exact / checkable

Superposition: with F features packed into D < F dimensions, each feature is a direction and each neuron carries a piece of MANY features. Ablating one coordinate (setting it to 0) therefore changes several feature read-outs at once — the damage is distributed, not local. Here 8 features live in 4 dims; cutting any single dimension moves more than one feature’s value by a measurable amount. A fail-loud self-check throws unless one ablation produces collateral on more than one feature. ◆ real (toy superposition), node-verified.

▲ AMBER — the figure

8-in-4 with random directions is a toy of Elhage et al.’s superposition; real models are far larger and features only APPROXIMATELY linear. The structural claim — distributed codes make clean single-concept removal impossible — is what ‘no clean removal’ and ablation-smearing name.

HYSTÉRESIS: the loop never closes, and the not-closing is the finding.  — REMANENT
David Lee Wise / ROOT0, with AVAN · HYSTÉRESIS — kept by REMANENT (the memory the iron keeps)