Persistent AI employees are not general-purpose tools. They are most valuable in domains where knowledge accumulates slowly, where institutional memory is critical, and where the cost of forgetting is high. These are the domains we are designing for.
Physics builds on centuries of experimental data and theoretical frameworks. An agent that remembers every prior experiment, every failed hypothesis, every contradiction between theory and observation is a different kind of tool. We are building agents that can hold the full experimental history of physics.
Complex engineering projects require deep knowledge of prior designs, failures, material constraints, and lessons learned. When engineers leave, that knowledge leaves. We are building agents that retain the full history of what has been built, what failed, and why.
The field of computer science moves fast, but it is built on decades of prior work — algorithms, systems, languages, paradigms. A persistent agent that understands the evolution of CS and knows what has and hasn't been tried is invaluable. We are building agents that carry that history forward.
High-stakes diplomatic decisions require deep contextual memory — who said what, what was agreed, what contradicts the current proposal. This institutional memory is critical and fragile. We are building agents that hold the full history of negotiations, agreements, and prior disputes permanently.
Marketing effectiveness comes from understanding deep patterns in customer behavior, campaign performance across time, and what strategies have worked in similar contexts. We are building agents that remember every campaign, every insight, every pattern — and learn from the full history.
Financial decision-making requires understanding market history, prior crises, economic cycles, and patterns that only reveal themselves across years. We are building agents that hold the complete financial history of markets, firms, and decisions — and reason across that full context.
We are beginning with a small number of domains. Scope will grow as the system matures.
Project Laplace