Belief 01
The next era is agentic. Copilots were a stepping stone.
Most of the value created by AI over the next ten years will accrue to systems that execute multi-step tasks autonomously, not to the chat interfaces that suggest them. Enterprises will buy outcomes, not assistants. We back the teams that have already crossed that line: founders building agents for legal work, financial operations, security, and the long tail of work the last cycle ignored.
What good looks like
Founders who can name the specific workflow their agent replaces, the measurable ROI it delivers, and the data they hold to keep the agent honest.
Belief 02
Infrastructure is the picks and shovels of the agent era.
Every company is becoming an AI company. Almost none of them have the rails to run AI safely, observably, or economically. The picks and shovels of this era are the tools that optimize models and hardware, and turn experimentation into deployment: evaluation, observability, inference efficiency. This is where the operating system of enterprise AI gets built.
What good looks like
Founders with deep developer empathy, a willingness to write boring documentation, and a worldview shaped by running infrastructure at scale.
Belief 03
AI is going vertical. Workflows beat models.
Foundation models are now commodities. The lasting companies will be built by founders who go deep into a single industry and earn proprietary data, deep integration, and category authority. Finance, legal, healthcare, insurance, logistics, manufacturing. The next Workday, the next Salesforce, and the next Bloomberg will be vertical AI operating systems. We are looking for the founders who already see this.
What good looks like
Founders with operating scars in the industry they are disrupting, an obsession with a single workflow, and a clear answer to why now.
Belief 04
AI governance and safety are not optional.
Every autonomous system creates risk. Agents act, machines decide, and someone has to answer for it. We back the security, governance, and safety layers that define what AI systems are allowed to do and verify what they actually did. Autonomy without control is not a product. It is a liability.
What good looks like
Founders who understand both AI and adversaries, who have shipped real products into hardened environments, and who take the question of control as seriously as the question of capability.
Belief 05
Intelligence is moving into the physical world.
The next wave of AI does not live on a screen. Robots, autonomous machines, and industrial systems that sense, decide, and act will reshape manufacturing, logistics, construction, and defense. The hardest problems sit exactly where models meet physics: real-time control, safety around humans, and hardware that has to work the first time. We back founders turning physical operations into autonomous systems.
What good looks like
Founders who have shipped hardware-plus-software products or run physical operations at scale, with a concrete answer for how their system behaves when the world does not cooperate.