Lightspeed

AI is upgrading the search engine. Lightspeed has backed the Exa team since we led their Series A raise in 2024. Today, we’re excited to announce that we’re participating in their $250M Series C, raised at a $2.2B valuation. Co-founders Will Bryk and Jeffrey Wang are building an API-first search engine for AI with Exa. Their search platform uses neural network architecture to help deliver higher-quality results. As AI products get smarter, their search functions need to evolve alongside them. Exa is building the infrastructure to help make that happen. CC: Lightspeed partners Guru Chahal and Nnamdi Iregbulem

America’s industrial base is at the precipice of a tune-up. Modern aerospace and defense systems demand faster iteration and stronger supply chains. Far too much critical infrastructure is currently dependent on fragmented production systems. That is why we are excited to participate in Amca’s $300M Series B at a $1B total valuation. Amca is building a new industrial platform that integrates engineering, testing, certification, and manufacturing into a single streamlined process. Through their growing nationwide factory network, Amca aims to dramatically reduce the time it takes to move production-grade hardware from development to deployment. Manufacturing speed is becoming a strategic advantage. We believe Amca is helping define what the next generation of industrial infrastructure can look like. Congratulations to Jai Malik, Eli Giovanetti, and the entire Amca team. Cc: Lightspeed Partner Connor Love

90% of successful cyberattacks start with phishing. As AI programs have evolved, email remains the primary entry point for enterprise attacks. And these advanced, AI-powered attacks continue to contribute to the high volume of phishing scams. Ocean is taking action by fighting back and building strong defense systems for enterprises. They’re leveraging AI agents that aim to investigate email in real time with human-level precision. Founders Shay Shwartz 🌊 and Oran Moyal may be in NYC and Tel Aviv, respectively, but they’ve come together to create this agentic email security platform. Today, they’re emerging from stealth with $28M in total funding. We can’t wait to see how their tech continues to expand and fortify their infrastructure. CC: David Gussarsky, Tal Morgenstern

You can’t secure what you can’t see. Modern enterprise infrastructure has become so vast that one of cybersecurity’s hardest problems is visibility. When Axonius launched in 2017, the pace of technological growth moved faster than many organizations could secure. The intent behind Axonius’ asset intelligence was to help companies finally see the full picture across devices, identities, apps, cloud infrastructure, and security systems, all from one place. Today, Axonius announced that it has reached $200M ARR and appointed Joe Diamond from interim to full-time CEO. Joe first served as Axonius’ CMO in May 2024, moved to the role of President in August 2025, and has been interim CEO since February 2026. Congratulations to Joe and the entire Axonius team. Lightspeed first invested in their Series C in 2020, and we are continuously excited by the mission and momentum.

One of the hardest questions to answer in regulated industries (Healthcare, Robotics, Defense) right now isn't whether or not to adopt AI, but how to act quickly while remaining safe and compliant. Last night, Ketryx hosted a dinner with visionaries across the healthcare sector, along with Lightspeed Partner Brenton Fargnoli, M.D. on these challenges. We're proud to back the leading team working to solve this critical issue. If you're wrestling with using AI in a regulated application, Ketryx helps teams deliver safer products, faster.

William Namgyal and Inigo Lenderking met as teenagers on the professional gaming circuit. Their friendship spanned time zones, William in the Bay Area and Inigo in Europe. Now they are building Luel. The two bonded over something specific: figuring out how to earn real income doing digital tasks. In their later teens, William and Inigo gravitated towards machine learning. They applied to UC Berkeley together, then dropped out together, just months later. Same team, new game. The idea at the core of Luel is the same one they bonded over as teens: everyday people, if given the right platform and opportunity, can do valuable digital work at a profound scale. Except now, the output is the training data powering frontier AI. We're so excited to co-lead Luel's $31.2M seed round alongside General Catalyst. Luel aims to be the global ground truth infrastructure for human data, where ordinary people all around the world get paid for turning elements of their daily lives into AI training data. 100+ countries, 1,000,000 submissions, and 40+ active dataset campaigns at any given time. AI can only be as good as the data it learns from. Luel turns human experience into AI training data, at global scale, on demand. We are thrilled to back them on this ongoing journey.

Last week, Ravi Mhatre sat down with Sebastian Mallaby at the Lightspeed SF office to discuss The Infinity Machine, his new book on Demis Hassabis, DeepMind, and the quest for superintelligence. Here were some of the topics Sebastian and Ravi explored last week: 1. In 1993, as a teenage chess prodigy who used to code until 4 AM and rejected a million-dollar offer to skip his education at Cambridge, Demis Hassabis decided his life would be devoted to building general intelligence. 2. What then set Demis apart was his insatiable curiosity and range, drawing from a vast career in neuroscience, gaming, philosophy, and physics. Sebastian made a compelling case for why Demis is one of the defining figures of modern AI. 3. Demis developed a “fluidity test” as a metric for competitiveness and longevity - part of the leader's job is to know the difference between ideas with potential to grow (those ideas being “fluid” and able to push forward) vs. those that lead to dead ends. He learned from his experience pushing a team too hard early in his career and ended up burning them out. Oversized ambition, lacking in delivery, was a recipe for disappointment. 4. What actually makes a frontier lab work: the toggle. Through trial and error, it became clear that the best labs aren’t operating in one constant mode. They just know how to balance patient, open-ended research and disciplined, tightly focused execution. 5. Build for the AI that exists in 6-12 months, not today. Ravi referenced a recent Dario Amodei quote about AI overtaking software engineering in 6-12 months, and asserted that it remains the most important mental model for founders building on top of AI. It should seem broken right now. The exponential improvement in model capability is deeply counterintuitive - it's invisible until it isn't. 6. The race that matters isn't models - it's deployment. Through his travels and research, Sebastian concluded that the real competition isn't who has the smartest model. Instead, those founders who embed AI most deeply into the real world may have the lasting edge. Deployment is the moat. 7. A bold idea that Demis has believed for years: information — not matter, not energy, not quantum states — may be the most fundamental primitive of the universe. Intelligence may ultimately come down to information processing. If so, the gap between biological and artificial minds may be smaller than we think. This was a fascinating discussion about where AI is actually headed. We are so grateful for Sebastian’s profound insights, Ravi’s discourse, and every person who joined us last week.

Human collaboration is often based on back-and-forth conversations. Interaction models let AI do the same. Many frontier AI models are made for long-running tasks that don’t keep humans in the loop, but Thinking Machines Lab believes the best work is done when humans and AI work together. They recently released a research preview of interaction models on their blog that goes into detail on the current turn-based trends, interactivity with AI, and their flagship model. Read the full article here.

The intersection of engineering and art holds immense potential. Flick sits at that junction. Co-Founders Ray Wang and Zoey Zhang built the AI-native filmmaking platform for creators who want to optimize the process of bringing their artistic vision to life. They’re also launching Flicker Residency, an initiative empowering filmmakers through mentorship and access to their technology resources. Flick just raised a $6M Seed Round, and we’re thrilled to be part of it. It’s inspiring to back artistic AI products that put craft and vision first.

Autoimmune and inflammatory diseases impact millions of lives, and yet there is still an enormous opportunity to rethink and improve the standard of care. Odyssey Therapeutics is on a mission to transform outcomes for patients who have been underserved by the current system, by targeting the immune system earlier, with the goal of stopping harmful inflammation before it starts. Odyssey is entering a new, important phase of development with clinical testing. The company has built a scalable drug discovery and development model and is advancing a diversified portfolio of programs to help deliver more effective medicines. By combining experienced pharma and biotech talent with a scalable drug development approach, Odyssey aims to deliver safer, more effective treatments to patients more efficiently. The company's lead drug, OD-001, is a RIPK2-targeted oral small molecule in Phase 2 for ulcerative colitis and Crohn's disease, with a Phase 2b readout expected in the second half of 2027. A second program, an SLC15A4 small molecule aimed at B-cell-mediated autoimmune diseases, is slated to enter Phase 1/2 trials in early 2027. Congratulations to Odyssey, founded by Gary D. Glick, on listing on Nasdaq as $ODTX and raising $279M from their IPO. We first invested in their Series D in 2025 and are thrilled to watch them embark on this new chapter.