ABOUT

LAST UPDATED: 2026-04-28

WHAT THIS IS

The AI Consciousness Tracker is an interactive, offline-first dashboard that aggregates publicly-available data on artificial intelligence capability, alignment, governance, and incident frequency into a single composite called the Signal Index (called the Consciousness Index in older documentation). It tracks eight threat vectors, ten dead-hand systems already embedded in critical infrastructure, six containment strategies and their relative readiness, and a literature archive spanning two centuries of fiction and philosophy that anticipated the present.

The site is free, contains no paywalled content, and the entire feature surface remains usable offline as a Progressive Web App. Source data is cited inline; methodology is disclosed at /features.html and inside the tracker itself.

WHAT IS REAL HERE // WHAT IS PRESENTATION

The Tracker is styled as a recovered terminal transmission. That styling — SIGNAL, [REDACTED], CLASS-7 CONTAINMENT, intercepted transmissions, glitch text — is voice, not classified intelligence. The underlying data and analysis are not fiction. This section is the contract:

Real

All 53 timeline events cite primary sources (AI Incident Database, arXiv, Nature, lab technical reports, Bloomberg, Europol, EU Commission records, Stanford HAI, OECD, etc.). Every event carries a SOURCES line.
The Signal Index formula, its five factor weights, per-factor uncertainty ranges, a ±10-point sensitivity analysis, and six academic citations (Bostrom 2014, Russell 2019, Ord 2020, Amodei et al. 2016, Hendrycks et al. 2023, Bengio et al. 2024) are all documented in the in-tracker METHODOLOGY panel.
The 34-work literature archive references real books. Eight public-domain works link directly to Project Gutenberg. The rest link to publishers or Amazon (affiliate).
The containment strategies (interpretability, corrigibility, formal verification, compute governance, behavioural monitoring, kill switches, boxing, alignment-tax reduction) map to real active alignment-research agendas.
The data pipeline runs on scheduled cron: benchmarks (SWE-bench, HumanEval, MMLU, GPQA, lab releases) refresh daily; the RSS / policy / incident-feed pipeline refreshes weekly. Every value carries a last-verified date.
The source code lives at github.com/Moonligh7er/AI-Consciousness-Tracker. The repository is currently private while we finish a security & secrets audit; it will be made public so anyone can inspect the methodology, reproduce the composite, and file issues.

Presentation

The Signal Index does not measure machine consciousness. Nothing yet can. It measures the conditions under which the question will be forced on us. See the honest note under the index itself.
Evidence tiers on timeline events (VERIFIED / CONSENSUS / ASSESSED / PROJECTED) are auto-assigned from source count and type — not hand-graded per event.
The Dead Hand section describes categories of embedded autonomous AI (algorithmic trading, credit scoring, facial recognition infrastructure, predictive policing), not named military systems. Each of the ten entries carries an expandable SOURCES list on the tracker, drawn from reputable public reporting: SEC/CFTC joint reports on flash crashes, DoD Directive 3000.09 on autonomous weapons, SIPRI, ProPublica, Nature, the NYT Clearview investigation, the US Consumer Financial Protection Bureau, and similar.
Default threat levels, initial slider positions, and the 78 % containment gap are editorial assessments. They are adjustable. The tool is the debate.

Corrections, source upgrades, and factual disputes are welcome via the feedback page. Updates ship with a visible changelog entry.

WHO RUNS IT

Moonlit Social Labs is an independent research and editorial outfit. The Tracker is its flagship project. We are not affiliated with any AI laboratory, government body, or commercial AI vendor; our independence is the point of the work.

Editorial decisions — what to track, how to weight it, when to revise — are made internally. Where we use external data, we name the source. Where we make a judgment call, we mark it as such in the methodology.

HOW WE WORK

Open methodology. Every weighting, threshold, and uncertainty range is documented in [How We Calculate the Consciousness Index](blog/how-consciousness-index-works.html). We disclose our priors.
Real citations. All blog articles use real, verifiable academic and journalistic sources. We do not fabricate citations or paraphrase secondary summaries as if they were primary research.
Steelmanned counter-views. Every analytical piece presents the strongest version of the opposing argument before reaching a conclusion. The Argument Builder tool inside the Tracker enforces this discipline procedurally.
Updates with timestamps. Every value carries a "last verified" date. Values that move are logged with reasons.
Offline-first. Nothing breaks if the network goes away. No feature depends on a third-party API for core function.

WHAT WE DO NOT DO

We do not sell data, run paid sponsorships, or accept directed funding from AI labs.
We do not pretend to forecast. The Consciousness Index is a present-state composite, not a probabilistic prediction.
We do not manufacture urgency. Where the data is reassuring, we say so. Where it is not, we say so.

SUPPORT

The Tracker is supported by display advertising and occasional reader contributions. If you would like to support the work directly, you can do so via Ko-fi. Supporter status hides advertising for that browser; see the Terms for details.

CONTACT

The fastest way to reach us is the Feedback page, which delivers messages directly to our inbox. For longer correspondence, including press, source-correction requests, and partnership inquiries, please use the same form and identify yourself in the subject line.

Editorial corrections, factual disputes, and source upgrades are welcome. We will respond to substantive correction requests and update the relevant pages with a visible changelog entry.

All Features — full list of what the Tracker does
Blog — long-form analysis
Privacy Policy
Terms of Use

COLOPHON

The Tracker is a static single-page application. No build step, no framework, no npm bundler on the client — every page is served as authored HTML, CSS, and JavaScript, so anyone with view-source can read exactly what runs. That transparency is part of the editorial posture.

Type

Inter for UI text and headings. IBM Plex Mono for labels, data, navigation, and metadata. Source Serif 4 for long-form essays and this page. Loaded from Google Fonts as a single stylesheet.

Palette

Near-black grounds with a blue cast (#0A0C10), warm paper-white ink (#E8E6E0), one accent — amber #E3B341 — used for the Signal Index, primary actions, and active nav. Semantic colours are all desaturated. Design thesis: declassified dossier meets mission-control instrument, not gaming HUD.

Stack

Front-end: vanilla HTML/CSS/JS. Canvas 2D for visualisations. Three.js for the singularity orb. Progressive Web App with service worker for offline use.
Back-end: Vercel serverless functions for the data pipeline (benchmark scrapers, RSS/policy feeds, keyword-heuristics engine). Upstash Redis for cached snapshots.
Cron: Vercel Cron. Benchmarks refresh daily; feed pipeline weekly.
Advertising: Adsterra banners, gated behind an explicit consent choice (see privacy policy). Amazon Associates for the literature archive.

Source

The source repository lives at github.com/Moonligh7er/AI-Consciousness-Tracker. It is currently private while a secrets-and-history audit is completed; it will be made public so anyone can inspect the methodology, reproduce the Signal Index composite, and file issues.

Licence

The editorial content (essays, methodology, timeline commentary, threat descriptions, ethics framework) is copyright © 2025–2026 Moonlit Social Labs; quotation and fair-use excerpting with attribution is welcome. The code is released under the MIT licence — see the repository LICENSE for the split between code (open) and editorial (all rights reserved).

MOONLIT SOCIAL LABS // SIGNAL INDEX v0.1 // PAGE REV 2026-08-29