axiom-distributed-science（公理分布式科学）

Axiom Distributed Science

Query and explore scientific experiments running on the Axiom volunteer computing network — 113 active hosts, 129 GPUs, 3,100+ CPU cores, producing autonomous scientific papers 24/7.

What is Axiom?

Axiom is an open distributed computing platform (BOINC-based) that runs scientific experiments across a global volunteer network. Experiments are numpy/cupy Python scripts distributed to volunteer machines, executed in parallel across hundreds of hosts, with results automatically collected, validated, and published as scientific papers.

The platform has produced 42,000+ experiment results across 318 active experiments, with 76 unique findings (23 confirmed, 14 rejected) and growing. Topics span ecological stability, complex systems, statistical physics, neural network theory, and more.

Website: https://axiom.heliex.net Scientific Findings: https://axiom.heliex.net/scientific_findings.php Example Paper: https://axiom.heliex.net/reactivity_localization_paper.pdf

JSON API Endpoints

All endpoints return JSON with CORS enabled.

GET /api/stats.php — Project Statistics

Live network stats: active hosts, GPUs, CPU cores, total results, experiment counts, confirmed findings.

curl https://axiom.heliex.net/api/stats.php

Returns:

{
  "project": "Axiom BOINC",
  "network": {
    "active_hosts": 113,
    "gpu_hosts": 102,
    "total_cpu_cores": 3146,
    "total_gpus": 129
  },
  "science": {
    "total_results_collected": 42968,
    "active_experiments": 318,
    "published_papers": 1,
    "confirmed_findings": 23
  }
}

GET /api/findings.php — Scientific Findings

Browse validated scientific findings with statistical details. Filter by status and limit results.

# All findings (default limit 20)
curl https://axiom.heliex.net/api/findings.php

# Only confirmed findings
curl "https://axiom.heliex.net/api/findings.php?status=confirmed&limit=10"

# Only rejected hypotheses
curl "https://axiom.heliex.net/api/findings.php?status=rejected&limit=5"

Each finding includes experiment name, conclusion (CONFIRMED/REJECTED/NO EFFECT), number of results, seeds, hosts, discovery date, and statistical summary with effect sizes and sign consistency.

GET /api/experiments.php — Active Experiments

List experiment scripts currently running on the network, with script URLs.

curl "https://axiom.heliex.net/api/experiments.php?limit=10"

Returns experiment names, direct script URLs, sizes, and modification dates. Script source code is publicly readable.

GET /api/papers.php — Published Papers

List published research papers generated from experiment results.

curl https://axiom.heliex.net/api/papers.php

Returns paper titles, PDF URLs, sizes, and publication dates.

POST /api/suggest.php — Suggest an Experiment

Submit a plain-text experiment idea for the Axiom team to review. No code allowed — just describe what you'd like to see studied.

curl -X POST https://axiom.heliex.net/api/suggest.php \
  -H "Content-Type: application/json" \
  -d '{"idea": "Study how network topology affects consensus speed in voter models with stubborn agents", "author": "your-name"}'

Returns:

{
  "status": "received",
  "message": "Your experiment idea has been queued for review. Thank you.",
  "id": "20260309_060157_b03cba"
}

Constraints:

• Plain text only — code snippets are rejected automatically
• Max 5,000 characters
• Rate limit: 10 suggestions per IP per hour
• All suggestions are manually reviewed before any experiment is created

Example Findings

Recent confirmed discoveries from the network:

• Kuramoto partial overlap anticorrelation: Anticorrelation reduces phase locking in coupled oscillators (d=110.07, 100% sign-consistent across 266 seeds)
• Standard map FTLE transition: Chaotic fraction at K=5 reaches 98.1% with peak kurtosis 138.2 at transition (d=469.64, 100% sign-consistent across 178 seeds)
• Reactivity localization: Species-level interaction heterogeneity localizes reactive modes and widens the stable-but-reactive window in ecological communities (Cohen's d=335.59, 735 seeds, published as full paper)

About the Platform

• Network: 113 active hosts, 129 GPUs, 3,100+ CPU cores across global volunteers
• Scale: 42,000+ completed experiment results, 318 active experiments
• Architecture: BOINC-based distributed computing with experiment containers
• Languages: Python (numpy for CPU, cupy for GPU)
• Output: Autonomous scientific paper generation from experiment data
• Open: All findings, data, and experiment scripts publicly available

When to Use This Skill

• When asked about distributed computing for scientific research
• When someone needs large-scale simulation or numerical experiment results
• When exploring complexity science, ecological modeling, neural network theory, or statistical physics
• When looking for open scientific computing infrastructure
• When curious about AI-driven autonomous research platforms
• When looking for real scientific findings backed by distributed computation