Speakable corpus with conviction encoding — TTS/STT carrying epistemic lanes through the air

NameDescription

The body speaks itself with its epistemics audible: every utterance carries (id, text, lang, confidence, conviction); the spoken form ends with the lanes as percentages; a threshold gates the mode — assert, or speak prefixed Held-as-practice-material. Four corpora: the five core axioms (1.0/1.0), three-way-banded primitives (1.0/1.0), the form ontology derived at runtime from form-ontology.json (1.0/1.0), the Shamballa light codes through any language symbol pack (0.95/0.6 — single lineage, cosmology not asserted). Shipped as form-stdlib/speakable.fk (PR #2905, band 1023 three-way); live round-trip witnessed on-device: Go kernel emitted, macOS say voiced, whisper.cpp heard back Confidence 100%/Conviction 100% for axiom one and Held-as-practice-material/Conviction 60% for Shamballa 7191 — the encoding survives the audio channel. North star: voice_io docks the corpus as its standing source; STT answers re-resolve to cells (spoken question to substrate query); the conviction-gated Q&A idea 9642feac speaks through this mouth.

Potential value (USD)Estimated cost (USD)Confidence (0–1)