A major milestone for American semiconductor manufacturing. Today, SandboxAQ announced a definitive agreement with the U.S. Department of Commerce for a $500 million CHIPS award to help tackle one of the most urgent challenges in American manufacturing: securing the critical materials and chemistries that underpin semiconductor production. Using Large Quantitative Models (LQMs), SandboxAQ will advance innovation across four material categories essential to the future of the industry: • PFAS-free process chemicals • Advanced catalysts • Rare earth-free magnets • Next-generation battery systems Across all four categories, the opportunity is the same: move critical materials from discovery to industrial deployment faster than ever before and strengthen the foundation of American semiconductor manufacturing. Read the full release here: sandboxaq.com/post/sandboxaq…

SandboxAQ Essential Concepts – SAIR: The Structural Context Behind Better Drug Discovery Models What if AI models had access to both measured drug potency and the 3D structure of the protein-ligand interaction? That’s the problem SAIR is built to address. SAIR brings together 5.24 million co-folded 3D protein-ligand structures with 1M+ experimentally measured, IC50-tagged protein-ligand pairs, giving AI models structural context that conventional potency datasets often lack. Here’s how it works: → Start with measured potency data from ChEMBL and BindingDB → Use AI to predict 3D protein-ligand structures → Generate multiple structures and computationally assess them → Release the structures alongside experimental potency, physical-validity checks, and model confidence The result is an open dataset designed to help models learn not just whether a molecule binds, but the structural context behind how it interacts with its target. SAIR is freely available for non-commercial use, with commercial access also available after a short form. Learn more with our infographic below and explore the dataset: sandboxaq.com/sair

What does it take to turn advanced sensing of the heart's magnetic field into a medical device and build the kind of team that can take it there? On the latest fAQ Podcast, Tai-Danae Bradley sits down with Kit Yee Au-Yeung, General Manager of AQ Med at SandboxAQ, to talk about building an AI-powered magnetocardiography device that is designed to capture the heart’s magnetic signals at the bedside, without the need for a cumbersome metal shield. Kit also gets into the less technical, and equally important, side of deep tech: leading through uncertainty, bringing quantum physicists into regulatory work, knowing when past playbooks no longer apply, and embracing the “tiger beetle” mindset. Tune in for a thoughtful conversation on building technology—and teams—at the frontier: piped.video/watch?v=_qHXuToM…

Heading to Highly-functional Material Week, MS&T26, or The Advanced Materials Show? Let’s talk about accelerating your materials R&D—from catalyst screening to alloys, batteries, and PFAS alternatives. 🚀 We'll be on hand to discuss the full ChemSim platform, including AQCat, our spin-aware catalyst model published in npj Computational Materials (nature.com/articles/s41524-0…). It delivers near-DFT accuracy up to 20,000x faster and uniquely accounts for spin polarization in key metals like iron, nickel, and cobalt. Catch our team on the floor: 🗓️ Sep. 30 - Oct. 2 | Highly-functional Material Week: Meet Deren Koseoglu to discuss the full ChemSim platform, including catalysts, batteries, PFAS alternatives, and alloys. 🗓️ October 4 - 7 (MS&T26) & October 6 - 7 (The Advanced Materials Show): Join Scott Healey, Tanner Kirk, and Sydnee King as we dive deep into alloys. See you there!

SandboxAQ Essential Concepts – Binding Affinity: The Molecular “Handshake” Behind Drug Discovery How do scientists know which drug candidates are worth making and testing? One important clue is binding affinity — a measure of how strongly a molecule binds to its biological target. A tighter molecular fit can mean a more stable complex, helping researchers identify the candidates most worth pursuing. But affinity is more than a number. It can help teams prioritize promising molecules, optimize designs, and de-risk candidates before they reach the lab. SandboxAQ combines AI with physics-based methods through its Large Quantitative Models (LQMs) to predict binding affinity, rank candidates, and help drug discovery teams search chemical space faster. The goal? Fewer guesses. Smarter experiments. Faster paths to promising drug candidates. Learn more with our infographic below 👇

SandboxAQ Essential Concepts – Drug Discovery: Small Molecules and Biologics Small molecules are exactly what they sound like: relatively small, chemically synthesized compounds. Their size allows them to cross cell membranes, making them especially useful for reaching targets inside cells. Many familiar medicines, from aspirin to statins and antibiotics, fall into this category. Biologics are much larger and more complex. Produced by living cells, they include antibodies, insulin, mRNA vaccines, and nanobodies. While their size generally keeps them outside cells, biologics can target cell-surface and extracellular proteins with high precision. Neither approach is inherently better. The biology of the target determines which modality makes the most sense, and some of the most difficult targets require new ways of designing and evaluating candidates. SandboxAQ’s Large Quantitative Models (LQMs) use AI grounded in physics to help design, screen, rank, and optimize candidates across both small molecules and biologics. The goal: expand what’s possible in drug discovery and find promising candidates faster. Small molecules and biologics are complementary tools. Choosing the right one is only the beginning. Learn more below.

SandboxAQ Essential Concepts – How Some Animals Sense Earth's Magnetic Field How do migratory animals navigate the globe without GPS? Part of the answer includes magnetoreception, a natural ability to read Earth’s magnetic signatures. Understanding how nature solves navigation challenges can also inspire new approaches to technology. It’s the same core idea behind SandboxAQ’s AQNav: using advanced sensing and physics to understand position and direction in environments where traditional navigation systems may fall short. Explore further in the infographic below!

We published a new blog post on the science behind AQCat, the model that treats spin polarization explicitly for iron, nickel, and cobalt at up to 20,000 times DFT speed. Those metals catalyze ammonia synthesis, which makes the world's fertilizer, and Fischer-Tropsch, which converts syngas into liquid fuels and chemical feedstocks. Both run at enormous industrial scale. Large public catalyst datasets tend to skip spin polarization because it costs far more compute, which leaves spin-unpolarized models wrong exactly where this chemistry lives. Omar Allam, Brook Wander, Sungyeon Kim, Aayush R. Singh and the catalysis team trained AQCat from scratch on the 13.5 million single-point DFT calculations in AQCat25, jointly with 20 million OC20 examples. That joint training gives the model its magnetic reach while it holds accuracy parity on general chemical space. npj Computational Materials published the peer-reviewed work, and we released the AQCat25 dataset, model checkpoints, and training code openly for non-commercial use. Read the full blog: sandboxaq.com/post/how-aqcat…

If you're excited about working on the frontier of positional intelligence and solving complex real-world problems. Come by booth 416 during AFA to meet some of our team! CTA: piped.video/-GxPYx3ssNQ #AFANational #positionalintelligence #quantumnavigation #altpnt

SandboxAQ Essential Concepts – Navigating Using Earth’s Crustal Magnetic Field Magnetic minerals in Earth’s crust create local variations in the planet’s magnetic field. These variations form patterns that can be mapped and used as a passive reference for positioning. The process is straightforward in principle: map the magnetic patterns, measure the field, and match the measurement to the map, but in practice, the magnetic map patterns can be hidden underneath an overwhelming amount of noise from the aircraft itself! That's where AQNav's AI-enhanced algorithms come in. By leveraging LQMs, we are able to separate the tiny signal from the noise. A magnetometer captures the magnetic field while navigation software compares the observed pattern against a magnetic anomaly map. Those matches can provide position updates and help bound drift in inertial navigation systems. This approach doesn’t replace existing navigation technologies. Instead, it adds another independent reference, using the Earth itself as part of the navigation system. Explore the infographic below to see how Earth’s crustal magnetic field can become a tool for navigation. 👇

Day one of AFA Air, Space & Cyber. Find us at Booth 416, to see how AQNav can ensure attritable platforms operate with confidence in contested, GPS-denied environments. #AFANational #positionalintelligence #quantumnavigation #altpnt sandboxaq.com/solutions/aqna…

SandboxAQ Essential Concepts – How Magnetometers Work A magnetometer detects the magnetic field at its location, turning an invisible physical phenomenon into a measurable signal. Different sensor technologies approach this in different ways, from Hall-effect and flux gate sensors to atomic vapor cells and nitrogen vacancy cells. Atomic magnetometers take an especially interesting approach: prepare atoms, let the magnetic field interact with their spins, and read the resulting response. But measuring the field is only the beginning. Real-world systems must separate useful magnetic signals from platform effects, environmental changes, and sensor noise, then combine those measurements with magnetic maps and other sensor data to estimate position. That measurement-to-mapping process is at the heart of magnetic navigation technologies like SandboxAQ’s AQNav. Explore the infographic below to see how magnetometers work—and how measurements of invisible magnetic fields can become actionable navigation data. 👇

Going to AFA? Booth 416 If you're in the DC area next week, come find SandboxAQ at Booth 416. We'll be showing how our new AQNav software delivers resilient, jam-proof navigation for autonomous platforms. sandboxaq.com/solutions/aqna… #AFANational #positionalintelligence #quantumnavigation #navigation #altpnt

SandboxAQ Essential Concepts – Why GPS Is Vulnerable GPS has transformed how we navigate, but its signals are faint, exposed, and vulnerable to disruption. Jamming can block signals. Spoofing can create convincing but incorrect positions. Buildings, terrain, and atmospheric conditions can distort reception, while space weather can affect how signals travel. So what happens when GPS becomes unavailable or untrusted? Reliable navigation requires resilience beyond satellite constellations. Independent sensors and alternative sources can help detect errors, maintain positioning, and keep systems moving when GPS is degraded or denied. That’s where technologies like SandboxAQ’s AQNav can play a role, using Earth’s magnetic signatures as a complementary reference alongside inertial systems and other navigation technologies. Explore the infographic below to see why GPS can be vulnerable, and how complementary navigation approaches can help. 👇

Heading to AFA Air, Space & Cyber this year? Find the SandboxAQ team at Booth 416 to see how AQNav's magnetic navigation software is delivering resilient, jam-proof positioning in GPS-denied environments. #AFANational #positionalintelligence #quantumnavigation #navigation #altpnt

SandboxAQ congratulates @NVIDIA on their public beta of BioNeMo Inference Runtime. We are proud to continue our collaboration with NVIDIA. Together, we are piloting a stack to make drug discovery both traceable and fast. Drug discovery turns on mechanistic questions, how a gene or pathway actually drives disease, and those answers only matter if a scientist can trace and defend every claim. We have built a stack that grounds every mechanistic answer in the AQ Knowledge Graph, and our pilot with NVIDIA aims to return it fast enough to use in conversation, served through NVIDIA BioNeMo Inference Runtime. The knowledge graph gives each claim a source: a node, an edge, a database, and its PubMed ID. BioNeMo orchestration and GPU-accelerated inference deliver the speed. A scientist asks a question in plain language and receives a fully traceable evidence pack, deployed where the data already lives, behind a partner's firewall or inside their cloud account. Explore BioNeMo Inference Runtime at: github.com/NVIDIA-BioNeMo/Bi… See SandboxAQ's work in drug discovery: sandboxaq.com/solutions/drug… Learn more at: sandboxaq.com/post/provenanc…

GPS goes down. The drone still has to know where it is. SandboxAQ CEO @jackhidary joined @CNBC's Power Lunch to talk through why AI systems need rigorous testing before they're deployed, and how magnetic navigation keeps autonomous systems oriented when GPS isn't there. Watch the full segment: cnbc.com/video/2026/09/09/ai…

Huge congratulations to @Flint_AI_ 's Switch for hitting #1 Product of the Day on @ProductHunt. It's a strong signal for bringing AI agents directly into the environments where teams already work, with shared context across Slack, Teams, Discord, and Telegram. Thanks to everyone who tried it, shared feedback, and helped us push Switch forward. Check out Switch on Product Hunt: producthunt.com/products/swi…

SandboxAQ has completed the world's first demonstration pairing quantum-enabled magnetic navigation with visual navigation on Northrop Grumman’s Group 3 attritable UAS. SandboxAQ engineers installed its AQNav software directly into existing systems of Northrop Grumman's Lumberjack® giving the platform GPS-free navigation. Read the full announcement to see what this milestone means for the future of combat drone operations. prnewswire.com/news-releases…

On leakage-controlled virtual screening, AQPotency delivers rankings statistically equivalent to physics-based docking, at a fraction of the time and cost. And because AQPotency doesn’t require a 3D crystal structure, it can reach targets that docking may not, including membrane proteins, mutant panels, and proteins that have never been crystallized. Every prediction comes with an uncertainty estimate and applicability score, giving scientists more than a ranking: a way to understand which results to act on and which to validate first. Join SandboxAQ’s Ben Shields, PhD, and Nihit Pokhrel, PhD, on October 7 to see AQPotency in action. 📅 October 7 | 11 AM PT Register here: sandboxaq.com/webinars/aqpot…

AI agents are moving beyond standalone interfaces. @Flint_AI_ ’s Switch puts agents directly into the collaboration environments teams already use, Slack, Teams, Discord, and Telegram, where they can access shared context, participate in conversations, and remain connected as the team evolves. Built to work with leading agent frameworks, including Claude Code, Google ADK, LangChain, and OpenAI, Switch is also open-source and free to use. Today, Switch is the number one Product of the Day on Product Hunt. See what Flint AI is building: producthunt.com/products/swi…