v0.2.0

Second tagged checkpoint of Orrin — an experimental symbolic-first cognitive runtime. ~318 commits since v0.1.1 (2026-06-15). Experimental research prototype; not production software and not a claim of sentience.

Highlights

Production, not just cognition. Reward is now denominated in durable outward effects via a content-addressed effect ledger (brain/agency/effect_ledger.py), with an evolving, human-ratified quality standard so production can't be gamed by template output. Aspirations are fail-able.

Goals & survival. A survival goal layer bridges autonomic pressure to cortical goals (preemption, recruitment, dormant tiers). Durable goals daemon with WAL/snapshots; surgical subgoal adaptation and reactive replanning.

Cognition. Pre-workspace symbolic binding and a decaying workspace-writeback loop; explore/exploit reach-value; adaptive learning-rate fixes; the LLM stays a gated, fail-closed, tool-only interface across pluggable providers (OpenAI / Anthropic / Gemini / local).

Existence model. Finite runtime-lifetime clock with phases and a terminal cycle; sealed .orrinlife run capsules; recoverable reset.

Product surfaces. Native desktop app (pywebview bridge, per-user data dir, OS-keychain secrets); the Face & Brain UI exposes the runtime as named rooms with a bilingual live thought line.

Engineering & docs. Biological-analogue terms removed codebase-wide; make verify gate; hermetic test suite (1,390+ tests); a full GitHub Wiki; green CI across ruff, mypy, pytest, coverage-ratchet, and the frontend build.

See the Wiki and `docs/ARCHITECTURE.md` for details.

v0.1.1

Orrin v0.1.1 — Windows build fix + I7 auto-update

v0.1.0

Orrin v0.1.0 — first cross-platform desktop build (unsigned)