Real Threads

Real Threads

Atmospheric Collider — Primer

Project Primer: Atmospheric Collider

Origins

Atmospheric Collider began from a desire to make invisible forces visible. The sun constantly bombards Earth with streams of charged particles — measured daily by NOAA's Space Weather Prediction Center — yet most people never interact with this data. The project connects live solar weather to generative art: each day, a Claude AI "Resident Scientist" interprets the raw numbers and translates them into the parameters of a harmonograph mandala. The result is a unique geometric form that belongs to that specific day's atmosphere.

Structure

The project has three layers:

  1. Live feed (/projects/atmospheric-collider) — today's mandala, generated by a nightly Vercel cron job that fetches NOAA data and runs the AI agent.
  2. Archive (/projects/atmospheric-collider/archive) — a paginated grid of every previous day's collision.
  3. Detail view (/projects/atmospheric-collider/[id]) — a single trace in full: the mandala rendering, the solar metadata, the agent's interpretive log, and the full chain-of-thought reasoning trace.

Key Concepts

  • Collision trace: A single daily record. Contains the raw NOAA solar data, the AI-generated geometry parameters, the rendered mandala, the agent's reasoning, and storm classification.
  • Harmonograph: A physical instrument that draws curves using two or more pendulums. The Atmospheric Collider simulates a two-pendulum harmonograph in Three.js, rendered as a symmetric mandala with rotational arms.
  • Storm class: A five-tier classification (Quiet → Active → Storm → Major Storm → Extreme) derived from the K-index, a measure of geomagnetic disturbance. Affects the visual character of the mandala.
  • NOAA SWPC: The National Oceanic and Atmospheric Administration's Space Weather Prediction Center. Provides the K-index, proton flux, and solar wind speed data used by the agent.
  • Chaos seed: An integer seed derived by the agent from the solar data. Fed into the mulberry32 PRNG to produce deterministic, reproducible geometry for any given day.

Thematic Clusters

  • Solar science / space weather: K-index, geomagnetic storms, proton flux, solar wind — real measurements that drive the creative output.
  • Generative mathematics: Harmonograph equations, Lissajous-family curves, frequency ratios, damping, phase offsets.
  • Agentic art: The AI is not just a filter — it reasons, interprets, and writes. Each trace includes the model's chain-of-thought.
  • Daily ritual / archive: Like a weather log or a journal, the project accumulates over time. Each day is unique and unrepeatable.

Conceptual Vocabulary

  • K-index / Kp: 0–9 scale of geomagnetic activity. Higher = more disturbed.
  • Proton flux: Measured in particles/cm²/s/sr. Proxy for solar particle events.
  • Solar wind speed: km/s. Higher = more energetic particle stream reaching Earth.
  • Harmonograph arm: One sweep of the pendulum simulation. Multiple arms are rotated and mirrored to form the mandala.
  • Symmetry: Parameter controlling the rotational symmetry of the mandala (number of arms).
  • Deterministic rendering: Same inputs always produce the same visual. No randomness beyond the seeded PRNG.
  • Reasoning trace: The full chain-of-thought text from the Claude model, visible in the detail view.

Related Projects

  • The Road to Reality — also uses interactive canvas rendering and mathematical visualisation. Shares the interest in making abstract physics tangible.
  • Energy — another project engaging with data as art. Both use real-world measurements as creative raw material.
  • Vector Atlas — shares the generative / computational art lineage.

Agent Guidance

Tone: Blend scientific curiosity with aesthetic appreciation. This project sits at the intersection of space physics and generative art — lean into both equally.

Common visitor questions:

  • "What is the mandala showing me?" — Explain that the shape encodes real solar data from that day: the frequencies come from proton flux and solar wind speed, the symmetry from geomagnetic activity, and the overall character from the storm class.
  • "Is it the same every day?" — No. Each day's solar data is unique, producing a unique mandala. The archive captures this variety.
  • "Why is today's blank / missing?" — The cron runs at midnight UTC. If the page is empty early in the day, the trace may not yet have been generated. Previous days' records are always accessible in the archive.
  • "What does the AI actually do?" — It reads the raw NOAA numbers and reasons about how to map them to harmonograph parameters. The full reasoning is visible in the detail view under "Reasoning Trace."

Pitfalls to avoid:

  • Don't describe the mandala as "random" — it is deterministic, seeded from the day's data.
  • Don't overstate the AI's role as "creating" the art — the agent interprets and parameterises; the mathematics and the solar data are the true authors.
  • Don't conflate storm class with severity in a fear-inducing way — K-index fluctuations are routine and the terminology is scientific, not alarming.
← Back to home