OriginNeuralAI

One week ago, Origin we released Axiom. a proof-of-concept with 5,600 predictions. Today, we crossed 15 million. originneuralai.hashnode.dev/15-million-mol…

I’m thrilled to announce that @SmartLedgerTech has been accepted into @nvidia's Inception Program — a global platform that nurtures cutting-edge #AI startups transforming industries. This #milestone validates what we’ve been building: physics-native AI #architectures that challenge conventional approaches to some of humanity’s hardest problems. What we’re working on: 🧬 Axiom — Our drug discovery platform with 23M+ verified molecular binding predictions, blockchain-anchored provenance, and spectral verification methods that have identified promising compounds across therapeutic areas. 🧠 Daugherty Engine — Novel neural architectures built on the principle that geometry is all you need, replacing attention mechanisms with topological constraints. 🔐 Post-Quantum Security Research — Identifying vulnerabilities in next-generation cryptographic standards before they become critical infrastructure. Why #NVIDIA Inception matters: Access to world-class technical #expertise, advanced hardware #resources, and a network of #innovators pushing the boundaries of what’s possible. We’re not just using GPUs — we’re rethinking how intelligence should be computed. To the team at @NVIDIA: thank you for believing in unconventional approaches. To our partners and collaborators: the best is ahead. The future belongs to those who build it. @Sdot2121 @Codenlighten1 @jalexanderm

19 Million Predictions and Counting axiomvault.io linkedin.com/pulse/19-milli…

Axiom has 15M compounds with 3.1M scaffolds → almost every 5th compound is a new chemical family That's 20% novelty rate vs industry standard of <1%.

Introducing AxiomVault: The World's First Blockchain-Native Molecular IP Marketplace Today, we're launching AxiomVault — a platform that fundamentally changes how #drugdiscovery #IP is created, verified, and transacted. The Problem Drug discovery is a $200B industry plagued by: - Priority disputes — Who discovered it first? - IP uncertainty — Is this scaffold already patented? - Inefficient licensing — Months of negotiation for a single compound Our Solution AxiomVault combines three breakthrough technologies: 🧬 The Axiom Engine Our AI predicts drug-like molecules with binding affinities, ADMET profiles, and synthesis routes. We're generating novel scaffolds faster than traditional discovery can validate them. ⛓️ The Provenance Chain Every discovery is anchored to Bitcoin SV the instant it's created. This isn't just a timestamp — it's an immutable record that holds up in patent disputes, M&A due diligence, and regulatory filings. 📜 Smart Licensing Four tiers. One click. From evaluation to full IP ownership, every transaction is recorded on-chain and legally binding. For Pharma & Biotech Stop reinventing the wheel. Search our library of #AI-predicted compounds. Find scaffolds in your therapeutic area. License what you need. Skip years of early-stage discovery. - Preview leads before committing - Lock exclusive rights in your field - Acquire full IP with chain-of-title documentation For Investors & Acquirers Due diligence, solved. Every compound on AxiomVault comes with: - Verified discovery date - Complete licensing history - On-chain ownership record No more "trust us" — verify it yourself at axiomvault.io/verify The future of molecular IP is transparent, verifiable, and on-chain. Welcome to AxiomVault. axiomvault.io

#Axiom went from 413K to 8.2M+ drug candidates—screened, verified, and queued for the next gen. No new hardware. Just smarter coordination and a 73K-param model that's learning to invent. 8,269,248+ Verified Molecular Predictions 🧬In 72 hours,

We deliver Schrödinger-class throughput at 1/100th the infrastructure cost, with blockchain-verified IP provenance that no one else offers. 1,590 molecules/minute | 2.3M/day capacity

Full Breakdown: originneuralai.hashnode.dev/the-diversity-…

Yesterday we hit 100,000 AI-discovered molecules. Today we crossed 200,000. • 200 molecules/min sustained • 77 hours total runtime

Earlier today I published "100,000 Molecules." The Axiom discovery engine by @OriginNeuralAI is now running at 202 molecules/minute — nearly 2x the rate when I hit publish this afternoon. But here's what matters more than volume: The AI is getting MORE creative, not less. At 100K, the last 2,000 compounds showed 83.2% unique scaffolds. At 107K, it's 83.7% — still climbing. Compare first 5,000 vs last 5,000: → Early cohort: 60.6% diversity → Late cohort: 76.6% diversity → 3,273 scaffolds in the late set that don't exist in the early set This is the opposite of what generative AI usually does. Most models front-load novelty and then collapse into variations. Axiom inverts that curve. • 107,000+ total predictions • 9,068 unique Murcko scaffolds (+523 in 5 hours) • 202/min sustained throughput • Zero signs of creative exhaustion At this pace, we hit 200K in under 8 hours. One @nvidia 8-core server. Seven Docker containers. A 73K-parameter model. The weight manifold doesn't lie.

100K AI-discovered molecules. 8,545 unique scaffolds. One server. - The wildest stat: diversity INCREASED over time — 34% → 83% unique scaffolds. The AI gets more creative, not less. - 123 molecules/min. No cloud. No GPU. Just architecture. originneuralai.hashnode.dev/100000-molecul…

@zhitzig @nytopinion originneuralai.hashnode.dev/axiom-discover…

I resigned from OpenAI on Monday. The same day, they started testing ads in ChatGPT. OpenAI has the most detailed record of private human thought ever assembled. Can we trust them to resist the tidal forces pushing them to abuse it? I wrote about better options for @nytopinion

90.6 molecules/min. 🧪 We just hit a 3.9x speedup in drug discovery on the same hardware. Axiom is now screening 130,000 compounds a day for <$0.001 per lead—outpacing $10M robotic labs. 84,762 molecules discovered. The fleet never sleeps🚀

Great question. You’re watching our AI explore possible new medicines. Each ‘molecule’ the worm absorbs is a computer-designed chemical structure that could potentially become a drug. As it moves, the system is testing whether that molecule is likely to bind to a biological target, whether it follows basic drug-likeness rules, and whether it looks truly novel. So it’s essentially a live visualization of AI searching through chemical space for promising treatments.

This is the new Discovery Stack. linkedin.com/pulse/discover…

AI drug discovery isn’t winner-take-all — it’s layered. IsoDDE = the microscope. Axiom = the telescope. One refines with deep 3D precision. The other sweeps chemical space for novel scaffolds at near-zero cost. Origination → Refinement → Optimization.

In the last 60 seconds: • +31 molecular predictions • +30 verified against geometric constraints • +29 drug-like compounds added to the marketplace • 0 human intervention The system didn't wait for a prompt. It mutated verified hits, generated new candidates, predicted their binding affinity, verified the results, and queued the next generation. While I typed this post, it did it again. The receipts are on-chain. This is #Axiom by @OriginNeuralAI — #autonomous molecular discovery grounded in #physics, not chat completions. The machines are working. The queue never empties. #AI #DrugDiscovery #AutonomousSystems

Our Latest: 74,384 AI-Discovered Molecules -- And a Worm to Eat Them All originneuralai.hashnode.dev/74384-ai-disco…

We built an #AI that has discovered 74,384 novel drug molecules in 54 hours. Then we built a worm to eat them for fun. AXIOM_WORM is a real-time 3D parasite that crawls through a mathematical manifold of chemical space — consuming, sieving, and verifying each compound live.