ATLASREALITY CALCULUS Enter the lab
METHODS / REPRODUCTION

Make the boundary
reproducible.

These checks validate the website and its local mathematical examples. They do not reproduce unavailable historical experiments.

Run locally

The website has no third-party runtime dependencies, accounts, or backend. Python 3 serves the static files. Node.js runs deterministic model checks.

python3 -m http.server 8000 --bind 127.0.0.1
# In another terminal, from the repository root:
python3 scripts/build.py --check
python3 scripts/verify.py
python3 reproduce/optical-check.py
node --test tests/core.test.mjs
node --check assets/app.js

The checked-in HTML works directly from a static server. Templates live in site/; the source inventory and register live in data/research.json. After editing those sources, run python3 scripts/build.py.

Optical comparison

The deterministic sweep evaluates 3,601 angles from 0° through 180° in 0.05° steps. It asserts a maximum absolute floating-point difference below 10⁻¹². Analytic equality follows independently from the double-angle identity.

Download optical-check.py

Transition explorer

A breadth-first search enumerates reachable states, including the start. Only admissible endpoints are traversed. The graph has no transition probabilities, experimental calibration, or time semantics. Tests cover cycles, blocked paths, observation equivalence, and invalid starting states.

Synthetic uncertainty

The laboratory uses fixed-seed normal deviates, sample standard deviation with n − 1 in the denominator, and the approximate interval x̄ ± 1.96s/√n. Independence and a normal-reference approximation are assumptions, not findings. The example excludes systematic error, sensor drift, and model misspecification.

Website acceptance

The browser suite exercises navigation, evidence filters, graph transitions, optical outputs, synthetic controls, offline boundaries, keyboard interaction, and responsive layouts at 390px and 1440px. It uses local Chromium through Playwright and axe-core as development-only tools. See the repository README for the exact commands.

Source and evidence policy

Supplied instructions are cited as project records, not substituted for original reports. The original complete v1.3 source and publications are unavailable / not independently reviewed. Historical software-test counts are not relabeled as current test results. No empirical sensor dataset is included.

Limits of validation

Automated accessibility checks and desktop browser emulation do not establish full WCAG conformance. A manual screen-reader audit, real-device testing, and independent scientific review remain open. ATLAS is now publicly hosted on Cloudflare Pages; the production HTTPS response headers, HSTS, framing protections and selected excluded source paths were externally checked on 29 October 2026. Future releases require repeated live checks after deployment. No independent penetration test is claimed.

Research source intake

Never import private research PDFs into a public build without editorial approval. A prospective import must record a document identifier, supplied original, source provenance, availability, version or checksum, allowed public distribution, and a deliberately assigned epistemic status. Importing a file does not prove its scientific assertions. The development repository records this procedure in docs/RESEARCH_IMPORT_POLICY.md.

Read all release limitations