We invest before the category is obvious.

How capable agents move from demos into teams: communicating across channels, navigating software, testing products, researching decisions, and acting within owner-defined boundaries.

Execution environments, authorization, observability, knowledge graphs, durable memory, data access, and local inference for agents that must work reliably over time.

What changes when intelligence is built into the operating layer of legal, finance, energy, health, science, restaurants, creative production, and enterprise data work.

How better models, automation, power systems, and operating software change aircraft, buildings, kitchens, manufacturing, ships, and the infrastructure around them.

What durable context looks like when people and agents need to retrieve, connect, and act on knowledge across long-running work.

Sandboxes, permissions, local inference, communication surfaces, and payment rails for software that must do real work rather than only answer questions.

Domain-specific systems that make legal, diligence, science, health, finance, and data work more legible without pretending judgment can be automated away.

The coordination, ownership, identity, and payment layers needed when employees, communities, software agents, and small brands participate directly.

New ways to make, edit, organize, and inhabit media when software participates in the creative loop and audiences help shape the result.

Products that bring health, aging, accessibility, and continuous support into homes and daily routines instead of leaving them inside institutional workflows.

Operational layers for the materials, machines, buildings, transport, and verification work required to turn physical decarbonization into repeatable systems.