abbe4becb6
Our own vocabularies in the styles of a growing core and a motor-plan set, built from cited open sources, never from a vendor's words, positions or names. Growing Core: one core board through eleven sizes (1x1 to 8x10). Every item lives at one point of the screen and is drawn, at every size, in the cell that contains it, so a word learned at 3x3 is in the same region at 8x10. The board is full at every size and nothing that appeared ever goes. The first boards are designed by hand; a solver places the rest by Fitzgerald region. Ten pages of more words, each with Back in the top-left corner; social messages with repairs; a Word forms page; hide-and-reveal at a fixed size from each word's stage. Fixed Paths: one meaning per location and one path per word, of 1 to 3 hits, at 30, 60 and 84 locations. The heaviest words (spoken frequency, more for Universal Core and Banajee 2003 words) take one hit in the easiest places for a blend of touch, scanning and gaze; a page is closed by the place of the key that opened it. No multi-meaning icon sequences. A vocabulary filter (starter) hides the rest in place. tools/pagesets: design sources (YAML), the solver (run once, committed as layouts), and a reproducible build that writes 19 .obz packages, an index, metrics and a per-language report to data/pagesets. aac-core tests/pagesets.rs imports every package with the app's own importer and checks positions, actions, word classes, that every symbol is the one the app draws, that each button says what the app's own export would, the growth and path rules, and language honesty. Co-Authored-By: Claude Opus 5.5 <[email protected]>
0 lines
Python
0 lines
Python