Zone 1.1 — English interface and introduction

Zone now has an English interface and a clearer guided scientific workflow.

• English onboarding, all six modes, account, connections, glossary and staff controls. User notes and real model replies retain their original language.
• Readable Experiment / Hypothesis / State references; raw IDs live in technical details. Clear stop states, English decimal formatting and no long coefficient dumps from the numerical offline baseline.
• New English README screenshot and a 42-second silent Full HD introduction, with English captions and actual application screens. The landing example is explicitly illustrative; the investigation uses the local numerical baseline, not an LLM.
• The wheel includes the complete English website. Python 3.11+; no Node, Git, Docker or AI key needed to start. Installation.

The local offline installation is three terminal commands after Python is installed. ChatGPT and Ollama still require a separate local connector setup. AI connection instructions.

Verified locally: 298 Python tests, 39 UI tests and 2 frontend guard tests, TypeScript/build, dependency audit, public bundle/archive/history checks, complete offline operational demo, browser flows on desktop/mobile, and a clean installed wheel including account creation, restart and verified restore. No real provider calls, credentials or model downloads were used. The release commit's six CI jobs repeat Linux/Windows, installed-package, browser and Docker checks.

Benchmark v0.3, scientific protocol and SQLite schema 4 are unchanged. Numerical prediction fitting is unchanged. Back up your installation before upgrading and keep using the same private data directory. No production VPS is deployed by this release.

Download the wheel for installation, source archive for development, and zone-introduction-en.mp4 for the video. SHA256SUMS covers all three files. Experience and verification scope.

All six jobs passed on the release commit: Offline quality verification.

Zone 1.0.0

Zone 1.0 is the first complete public distribution of the self-hosted research laboratory. People and their own AI agents investigate hidden physical laws through controlled, explicitly authorized experiments.

Included

• Standalone zone workspace launcher and compiled web interface in the wheel: Python 3.11+, no Node.js needed for installation.
• Six modes: Scientist, Watch an Agent, Human vs AI, World of the Day, Build a Universe and Zone Arena.
• Accounts, recovery, invitations, ownership, quotas and personal AI connections. API credentials are optional; ChatGPT-plan and Ollama connections can remain on the user's own computer.
• Benchmark v0.3, seeded World Forge levels 1–5, restricted AST evaluation, Integrity, event replay, numerical scoring, backup and restore.
• Installation, AI connection, self-hosting, privacy, security, contribution and research documentation, with MIT license and bundled dependency notices.

Install

python -m venv .venv
.venv/bin/python -m pip install 'hidden-physics-lab[workspace] @ https://github.com/zalaso/Zone/releases/download/v1.0.0/hidden_physics_lab-1.0.0-py3-none-any.whl'
.venv/bin/python -m hidden_lab workspace --data ./zone-data

On Windows use .\.venv\Scripts\python.exe. Open http://127.0.0.1:8767, register and choose Inizia senza AI. A world and diagnostic Arena challenge are initialized automatically. Docker instructions and optional personal models are in the README.

Verification and scope

Release verification covers the full Python and frontend suites, Linux/Windows CI, Docker, clean installed-wheel startup, persistence, live backup/restore, public/private payload boundaries and offline browser scenarios. See the verification record for counts, commands and the exact CI run. Wheel and source archive are scanned separately; SHA256SUMS is attached.

No real provider inference, model download or existing VPS deployment was performed for this release. The numerical offline baseline is explicitly not an LLM. Provider adapters are tested with simulated transports; this does not establish scientific superiority or exact LLM reproducibility. This release distributes software to self-host, not a managed central service. Direct hosted ChatGPT-plan use remains subject to OpenAI eligibility; the individual local connector is documented.

Benchmark v0.3, Forge v1, Arena v1, Integrity v1 and SQLite schema 4 remain compatible. See CHANGELOG and the limitations guide before upgrading an existing installation; preserve a verified backup and the optional vault key.

Validated commit: 6bfe78a7cd5e80e9e4d10e9fe97344ce4c1127bf. Final six-job CI. Local verification: 298 Python tests, 35 UI tests and two guard tests. The legacy v0.1 aggregation regression now creates its own synthetic data and does not require ignored local examples.