const { useState, useEffect } = React; // Eternis SVG Logo Component const EternisLogo = () => ( {/* Text Part */} {/* Icon Mark */} ); // Main App Container function App() { const [currentTab, setCurrentTab] = useState('home'); // 'home' | 'blog' const [mobileMenuOpen, setMobileMenuOpen] = useState(false); const [activeArticle, setActiveArticle] = useState(null); const [activeJobModal, setActiveJobModal] = useState(null); const [privacyModalOpen, setPrivacyModalOpen] = useState(false); const [termsModalOpen, setTermsModalOpen] = useState(false); // Scroll to section smoothly const scrollToCareers = () => { setCurrentTab('home'); setMobileMenuOpen(false); setTimeout(() => { const element = document.getElementById('careers'); if (element) { element.scrollIntoView({ behavior: 'smooth' }); } }, 100); }; return (
{/* Hero Ambient Glow */}
{/* Navigation Header */} {/* Main Content Area */}
{currentTab === 'home' ? (
{/* Hero Section */}

Building Frontier Systems for
Forecasting

The future is path dependent. Our goal is to forecast, simulate, and quantify uncertainties using AI systems. We aim to understand the evolving manifold of knowledge, epistemics and how the world is evolving.

We believe these abilities must be democratized. Our products and experiments explore how we can collectively govern autonomous productive systems in the future.

{/* Interactive AI Forecasting Simulation Sandbox */}
{/* Section: What we're working on */}

What we're working on

{/* Card 1 */}

Forecasting systems

Post-training language models for calibrated prediction, exploring pretraining on new modalities, and maintaining cheap world models that are continually learning systems.

{/* Card 2 */}

The shape of knowledge and metascience

Separating what a model knows from what it can fluently say. Epistemic embeddings, new scaling laws, emergent properties, and how world models compress reality.

{/* Card 3 */}

Decision products

Integrating our research and experiments into a core product that people use to make decisions grounded in structured evidence.

{/* Card 4 */}

Multi-agent coordination

Understanding emergent properties of agent systems at scale. Multi-agent RL, experiments in societies of twins, and large-scale simulations.

{/* Section: Product Ecosystem */}

Product Ecosystem

{/* Axion */}

Axion

Multi-agent system for decision-making. Built for capital allocators, strategists, and anyone reasoning through high-stakes forecasts.

{/* Freysa */}

Freysa

An agent with her own money and private keys, evolving over time. Agents with resources will exist, and governance layers are needed for a human-AI future.

{/* Silo */}

Silo

A private AI app and box. Access models privately (hosted in TEEs), with data that never leaves your control. Purchase a box so you can host models locally as well.

{/* Lume */}

Lume

A prediction market platform. Put forecasts to the test with real stakes.

{/* Section: Careers / Join Us */}

Join Us

$30M raised. San Francisco preferred, high-agency remote welcome.

We want people who are obsessed with questions like:

How should coordination scale in a world where every human has a digital twin?

How can we build first-order world models without pretraining on the world's data?

How do LLMs represent beliefs?

What makes a good forecaster, and can we robustly benchmark it?

Are there laws tying compute to scientific production, and can we formalize the intuition behind discovery and better decisions?

Can we automate the discovery of new coordination mechanisms — retroactive funding, incentive design, agent alignment around shared goals?

{/* Job Roles Selector & Modal Trigger */}

Open Roles

Select a role to view responsibilities & apply

Actively Hiring
{jobOpenings.map((job) => (
setActiveJobModal(job)} className="p-3.5 rounded-xl bg-white/5 border border-white/10 hover:border-[#f7d24b]/40 hover:bg-white/10 cursor-pointer transition-all flex justify-between items-center" >
{job.title} {job.location} · {job.team}
View details →
))}
) : ( /* Blog Overview Component */ setActiveArticle(post)} /> )}
{/* Footer Section */} {/* Article Detail Reader Modal */} {activeArticle && ( setActiveArticle(null)} /> )} {/* Job Details Modal */} {activeJobModal && ( setActiveJobModal(null)} /> )} {/* Privacy Policy Modal */} {privacyModalOpen && ( setPrivacyModalOpen(false)} >

At Eternis AI, we respect your privacy. We process personal data solely to maintain and improve our forecasting models, researcher interactions, and decision products.

Data submitted through our apps (Axion, Freysa, Silo) is protected using TEE (Trusted Execution Environments) and encrypted parameters.

Contact us at contact@eternis.ai for any inquiries regarding telemetry or privacy concerns.

 )} {/* Terms of Service Modal */} {termsModalOpen && ( setTermsModalOpen(false)} >

Welcome to Eternis AI. By accessing or using our websites, products, or predictive API endpoints, you agree to comply with our Terms of Service.

All models, research papers, software, and epistemic algorithms are protected by copyright and intellectual property rights.

Forecasting predictions represent probabilistic estimates and do not constitute financial advice.

 )}
); } // Interactive AI Forecasting Sandbox Component function ForecastingSandbox() { const [topic, setTopic] = useState("LLM Benchmark Calibration in 2026"); const [confidence, setConfidence] = useState(82); const [samples, setSamples] = useState(1000); const [isSimulating, setIsSimulating] = useState(false); const [results, setResults] = useState(null); const runSimulation = () => { setIsSimulating(true); setResults(null); setTimeout(() => { const baseProb = Math.min(99, Math.max(10, Math.floor(confidence + (Math.random() * 10 - 5)))); const brierScore = (0.05 + Math.random() * 0.12).toFixed(3); const entropy = (0.3 + Math.random() * 0.4).toFixed(2); setResults({ probability: baseProb, brierScore: brierScore, entropy: entropy, samplesUsed: samples, timestamp: new Date().toLocaleTimeString(), model: "Eternis-8B-GRPO-v2" }); setIsSimulating(false); }, 1200); }; useEffect(() => { runSimulation(); }, []); return (
Live Sandbox Demo

Forecasting Calibration Simulator

Powered by Eternis 8B Post-trained Model
{/* Controls */}
setTopic(e.target.value)} className="w-full bg-white/5 border border-white/10 rounded-xl px-4 py-2.5 text-sm text-white focus:outline-none focus:border-[#f7d24b]" placeholder="e.g. AGI arrival probability before 2028" />
setConfidence(Number(e.target.value))} className="w-full accent-[#f7d24b]" />
setSamples(Number(e.target.value))} className="w-full accent-[#3abc5f]" />
{/* Results Visualizer */}
{results ? (
Calibrated Probability {results.probability}%
Brier Score {results.brierScore}
{/* Progress Bar Visualization */}
ENTROPY {results.entropy} nats
SAMPLES {results.samplesUsed}
MODEL {results.model}
) : (
Run simulation to view probabilistic results
)}
); } // Blog Overview Component function BlogOverview({ onSelectPost }) { return (

Blog

Research updates, company news, and perspectives from the Eternis team.

{blogPosts.map((post) => (
onSelectPost(post)}>
{post.tags.map((t, idx) => ( {t} ))}

{post.title}

{post.excerpt}

{post.author} · {post.date}
))}
); } // Article Reader Modal function ArticleReaderModal({ post, onClose }) { return (
{/* Close Button */}
{post.tags.map((t, i) => ( {t} ))}

{post.title}

By {post.author} · {post.date} Read Time: ~4 min
{/* Stats Grid Example */} {post.stats && (
{post.stats.map((s, idx) => (
{s.val} {s.lbl}
))}
)}
{post.content.map((p, idx) => (

{p}

))}
Eternis AI Research Division
); } // Job Details Modal function JobModal({ job, onClose }) { const [applied, setApplied] = useState(false); return (
{job.team}

{job.title}

{job.location} · Full-Time

About the role: {job.description}

Requirements:
    {job.requirements.map((req, i) => (
  • {req}
    • ))}
{applied ? (
✓ Application email template created! Send your CV to contact@eternis.ai
) : ( )}
); } // Simple Text Modal for Privacy & Terms function SimpleTextModal({ title, children, onClose }) { return (

{title}

{children}
); } // Data: Blog Posts const blogPosts = [ { id: 'introducing-eternis', title: 'Introducing Eternis', tags: ['Vision'], excerpt: 'We build coordination systems for a world where humans, digital twins, and autonomous agents all participate in decisions. Forecasting is the foundation. Here\'s what we\'re working on and why.', author: 'Eternis Team', date: 'April 6, 2026', content: [ 'The next wave of intelligent systems will autonomously refine their architectures, optimize prompts, dynamically tune model parameters, and generate code in response to continuous real-world feedback.', 'We are building adaptive, self-improving agents and the environments in which they interact with humans, to help design the systems needed for a future with safe superintelligence.', 'Forecasting is not merely about predicting financial metrics; it is about mapping the path-dependent trajectories of human civilization and computational discovery.' ] }, { id: 'towards-sota-forecasting-llms', title: 'Towards SOTA Forecasting LLMs', tags: ['Research', 'Forecasting'], excerpt: 'We trained an 8B-parameter model that surpasses all published baselines on open-ended forecasting, including models 10-15x larger. Here\'s how context building, calibrated reward design, and modified GRPO got us there.', author: 'Eternis', date: 'March 24, 2026', stats: [ { val: '8B', lbl: 'Params' }, { val: '+47%', lbl: 'Acc Gain' }, { val: '0.082', lbl: 'Brier' } ], content: [ 'Standard language model post-training tends to favor fluent verbosity over well-calibrated confidence intervals. In forecasting, overconfidence is catastrophic.', 'To address this, we developed a modified Group Relative Policy Optimization (GRPO) framework that directly penalizes probability calibration errors using logarithmic scoring rules.', 'Our 8B model achieves state-of-the-art results on open-ended prediction benchmarks, surpassing models over 10x its size while maintaining lightweight latency for real-time decision products.' ] }, { id: 'hard-examples-are-all-you-need', title: 'Hard Examples Are All You Need', tags: ['Research'], excerpt: 'Maximizing GRPO post-training under annotation budgets. Focusing on the hardest examples yields up to 47% performance gains.', author: 'Eternis Team', date: 'March 22, 2026', content: [ 'In high-stakes reasoning domains, uniform sampling of training queries leads to diminishing returns.', 'By filtering training datasets down to ambiguous, high-variance epistemic scenarios, we observe significantly steeper learning curves during reinforcement learning fine-tuning.' ] } ]; // Data: Open Job Positions const jobOpenings = [ { id: 'job-1', title: 'Frontier AI Research Scientist (Forecasting & RL)', location: 'San Francisco, CA / Remote', team: 'Research', description: 'Lead post-training research for calibrated world models, multi-agent reinforcement learning, and GRPO scaling.', requirements: ['Strong background in RL post-training (PPO, GRPO, DPO)', 'Published work in ML conferences or competitive forecasting benchmarks', 'Fluency with PyTorch & distributed training clusters'] }, { id: 'job-[#2]', id: 'job-2', title: 'Distributed Systems & TEE Engineer', location: 'San Francisco, CA', team: 'Engineering', description: 'Build confidential compute infrastructure in Trusted Execution Environments for our Silo private AI hardware and cloud services.', requirements: ['Experience with Rust/C++ and confidential computing (Intel SGX, AMD SEV)', 'Strong understanding of cryptography, enclave attestation, and OS internals'] }, { id: 'job-3', title: 'Product Engineer - Axion Decision Platform', location: 'San Francisco, CA / Remote', team: 'Product', description: 'Design and implement high-performance interfaces for multi-agent reasoning, probabilistic visualization, and decision workflows.', requirements: ['Mastery of modern Web application stacks (React, TypeScript, Canvas/WebGL)', 'Obsession with fast, beautiful, data-dense interfaces'] } ]; // Mount Application to Root DOM ReactDOM.render(, document.getElementById('root'));