Observe with context.

01 / What is AREAX?

AREAX is a fullstack prototype for exploring sky anomaly research. UFO refers to an unidentified object, not a claim of extraterrestrial origin. This version includes synthetic telemetry, statistical novelty detection, persistent observations, and a review interface.

All observations originate from simulation. No live camera, radar, satellite, or tracking service is connected. The “LIVE FEEDS” panel presents animated fictional scenes, each marked SIM.

02 / System architecture

The React frontend uses Next.js. API routes run in the Node.js runtime on Vercel. Drizzle ORM connects to a persistent Turso/libSQL database. Filters and visual animations run in the browser; observation and review records are stored on the server.

Each scan generates three synthetic features, scores their deviation, then saves the result with its UTC timestamp, region, reproduction seed, source, and model version. The API only reports success after the database insert completes.

A monochrome orbital saucer SVG is shared by the header logo and favicon. The map combines an AI-generated texture with interactive SVG markers. Control-room and feed imagery is fictional AI-generated artwork animated with CSS layers.

03 / Model & methodology

AREAX uses a diagonal Gaussian novelty detector fitted to 512 synthetic baseline samples with the fixed seed 42. This is a simple statistical anomaly model, not a computer vision system or a trained UFO classifier.

d = sqrt(Σ ((xⱼ − μⱼ)² / σⱼ²))
novelty_score = round(min(100, d / 8 × 100))
ANOMALY if novelty_score ≥ 65

About 35% of generator calls use wider feature distributions; these do not always cross the anomaly threshold. The threshold of 65 and distance scale of 8 are demonstration choices, uncalibrated against real observations. A score of 90 does not mean a 90% probability of a UFO.

Visual scenes are independent of the numerical features analyzed. No objects or features are extracted from images. Record details preserve the seed for replay through simulate(seed), the feature vector, statistical distance, and per-feature contributions. The storage status ANOMALI is retained for compatibility and displayed as ANOMALY in the English interface.

04 / Data provenance

• Telemetry: internal pseudorandom generator, seeded using server-side Web Crypto.
• Live Feeds: AI-generated still artwork with animated camera drift, mist, light trails, rain, and water reflections. These are simulated scenes, not camera streams.
• Regions: four geographical examples for demonstration. Their inclusion does not imply real instruments or sightings at those locations.
• Map: illustrative AI-generated texture, approximate positions, and schematic scale. Not suitable for navigation.
• Typography: Google Fonts with system font fallbacks when unavailable.

Real source integration is not implemented. It requires source permissions, sensor calibration, provenance records, time synchronization, and comparisons with aircraft, satellites, weather, birds, and camera artifacts. Changing a label cannot enable real monitoring.

05 / Operating the observatory

1. Select a map marker or region in the map sidebar to change the scan target. Clicking a feed illustration also selects its associated virtual region and opens the analysis console.
2. Select “Run simulation” to generate, analyze, and save one observation.
3. Filter by All signals, Anomalies, or the active region.
4. Open a record using the row arrow or detection card. Inspect its features, score, UTC timestamp, model, and seed.
5. Mark records as reviewed or undo the mark. This is a workflow annotation, not confirmation of a phenomenon.
6. Use the download icon to export up to 200 latest observations as JSON.

The table shows up to 20 filtered rows. Summary counts cover the latest 200 loaded records rather than the entire database. “Pause motion” or the feed pause button stops scene motion and the elapsed display together. Reduced-motion preferences are respected automatically.

The X icon links to the official AREAX profile.

06 / API reference

Valid regions: nevada, atacama, iceland, lombok. Errors: 400 invalid input; 403 cross-origin request; 404 record not found; 413 oversized body; 503 unavailable database or failed persistence.

Configure access protection for your Vercel deployment before granting access. The application does not implement accounts or tenant isolation: authorized visitors share one dataset. Protect independent deployments with an authenticated gateway. An Origin check is not authentication.

07 / Run & deploy

The downloadable source includes a README, dependency lockfile, configuration, SQL migrations, and .env.example. Use Node.js 22.13+ and npm. Local and hosted databases are separate.

npm ci
npm run build
npm run db:migrate
npm start

Open the loopback URL printed by the server. For development with hot reload, use npm run dev after applying local migrations. For Vercel, select the Next.js framework preset, use npm run build, and leave Output Directory at the framework default. Set TURSO_DATABASE_URL and TURSO_AUTH_TOKEN for the remote database, then apply migrations before using the API.

The README explains Vercel deployment and remote database setup. Local file: URLs must never be used on Vercel. Builds do not require database credentials; scans do. The official X profile is set in lib/socials.ts.

08 / Know the limits

• This prototype cannot detect, prove, or classify real UFOs.
• The model learns only a synthetic baseline. Real-world accuracy, false positives, and generalization have not been measured.
• Feature independence and Gaussian distributions are simplifying assumptions; the baseline generator itself is uniform.
• No scheduled scanning, live ingestion, automatic retraining, or physical sensors are connected. Scans run on demand.
• A high score indicates deviation, not a cause. Ordinary objects and sensor artifacts may create anomalies.
• There is no application rate limit, automatic retention, or reviewer identity audit. Keep this prototype private.

Use AREAX to explore interface and data workflows. Scientific conclusions require verified data, calibrated methods, and independent review.