At the recent All In conference, the emerging and secretive world of AI world models took center stage, highlighting two notable companies: AMI Labs led by Yann LeCun and World Labs under Fei-Fei Li. Both have attracted significant funding and attention but have been notably tight-lipped about their progress and product plans. World models, which aim to automate spatial intelligence, hold promise for applications across robotics, interactive media, and autonomous vehicles; however, concrete commercialization steps remain unclear. Michael Rabbat, AMI Labs’ co-founder, emphasized their ongoing research phase and reluctance to share specific product details publicly.

World Labs’ most advanced product, Marble, showcases capabilities like media creation and interactive gaming environments but also refrains from disclosing definitive use cases or commercialization timelines. This culture of secrecy extends beyond the companies themselves to their data suppliers. For example, Physicl’s CEO Alex de Vigan expressed frustration at not knowing how his firm’s data is being applied, despite its acknowledged usefulness. This confidentiality underscores how versatile and potentially lucrative world models are, ranging from simple navigable maps to complex humanoid robotics, making competitive secrecy a strategic priority.

The vast potential of world models has prompted companies like AMI to explore diverse fields such as manufacturing, biomedicine, and healthcare software, often through partnerships like AMI’s collaboration with Nabia. Yet the ambition to cover multiple avenues is tempered by a common strategy to avoid revealing too much too soon. Announcing a breakthrough or product prematurely risks attracting competitors, including other startups, new AI labs, and industry giants like OpenAI and Anthropic. For now, these firms are using abundant venture funding to build cautiously behind the scenes while maintaining a shroud of mystery.

This atmosphere reflects a classic “dark forest” scenario in competitive tech: staying quiet is essential to avoid drawing attention that could jeopardize hard-earned advantages. While the world models sector is flush with cash and excitement, the race remains tightly contested and invisible to outsiders. Companies preserve their lead by controlling information tightly, even from their collaborators. This approach keeps the industry in suspense about when and how world models will move from research labs into impactful real-world applications.