The first public release of Altar — a provider-neutral, moment-selected symbolic field for AI.
What it is. Altar introduces a reproducible symbolic difference into an AI interaction. An explicit UTC moment and a frozen, content-addressed symbol pack select coordinates before interpretation; the question never enters the selector.
Highlights
- Deterministic selector
altar-portable-v2with exactly five inputs: selector_version · pack_sha256 · unix_milliseconds · domain · counter whole-v3canon: 32 symbolic systems, 671 coordinates, system-balanced sampling- Auditable receipts with geometry and selection proofs; checked-in golden vectors for independent implementations
- Geometries: Silence · Note · Chord · Field · Tune, plus an auto gate frozen at 85/12/3
- Live Thinking 3/6/9 — depth is chosen first; each next symbol is drawn only after the previous contact
- Modern Dreamspell day context, computed after selection (not the traditional Maya calendar)
- Three integration routes: self-contained AI skill, JSON CLI, Python package — zero runtime dependencies, no network calls, no telemetry
- Gene Keys carried as numbered keys with their public I Ching hexagram correspondence only
Claim boundary. Mechanism and reproducibility are verified; correspondence and answer influence remain open, testable hypotheses. See VALIDITY.md.
Quick start
pip install git+https://github.com/light-language-models/altar
altar-ai field --now --timezone UTC
License: Apache-2.0.