Turing Research is launching a groundbreaking initiative to capture and utilize the complete, unfiltered operational history of companies, creating the definitive dataset for training the next generation of frontier models. Project Lazarus is an initiative to acquire and permanently preserve the full, unfiltered operational history of defunct or inactive companies at scale. We focus on private codebases, version histories, internal documentation, post-mortems, experimentation logs, infrastructure tooling, and everyday work artifacts that collectively reflect how real organizations actually operate. These materials capture the reality of knowledge work: incomplete specifications, tradeoffs made under time pressure, accumulated technical debt, evolving systems, and decisions made under uncertainty. Unlike polished outputs, operational traces preserve the causal structure of work across weeks, months, and years. We prioritize industries with high complexity and outsized GDP impact, including financial services, healthcare and pharma, advanced manufacturing, and enterprise software. These domains contain long-horizon decision making, regulatory constraints, supply chain dependencies, and high-value intellectual property that are critical for training economically useful AI systems. The data is structured for advanced methodologies such as reinforcement learning, imitation learning & long-horizon task evaluation, enabling models to learn multi-step reasoning, organizational decision processes, and system diagnosis over extended timelines. For founders, Project Lazarus is also preservation. A company’s history is a compressed record of human judgment, experimentation, and problem-solving. Instead of disappearing, that work compounds by becoming part of the foundation shaping the next generation of autonomous AI systems.