The Agentic Thesis
Why the next decade of software belongs to systems that think, decide and act — and what that means for how we build ventures.
For most of software’s history, we built tools. A tool waits. It sits there, perfectly inert, until a human reaches for it and pulls a lever. The better the tool, the smoother the lever.
Agentic systems break that contract. They don’t wait — they observe, reason and act on your behalf, inside guardrails you define once and trust thereafter. That shift sounds small. It is not. It changes what a product is.
From features to faculties
A feature is something a product can do when you ask. A faculty is something a product does because it understands the goal. When AgentPay Cards provisions a virtual card, screens it for fraud and reconciles the ledger without a human in the loop, those aren’t five features stitched together. They’re one faculty — handle money responsibly — expressed through autonomous agents.
The design question stops being “what buttons do we ship?” and becomes “what judgement do we delegate, and how do we keep it accountable?”
Three principles I build by
- Delegate judgement, not just labour. Automation moves work. Agency moves decisions. The leverage is in the second one.
- Guardrails before autonomy. An agent is only as trustworthy as the boundaries around it. Define the box first; then let it run free inside.
- Make the invisible legible. When systems act on their own, the interface’s job is to show why. Trust is a design surface.
The goal isn’t software that does more. It’s software that needs you less — without ever surprising you.
What this means for ventures
Every venture in the portfolio is a bet on the same future: that intelligence becomes infrastructure. VYLIOX orchestrates agencies. VICTTORY surfaces talent. REPULIZE defends reputations in real time. Different domains, one spine — agentic systems doing the work a roomful of people used to.
We’re early. But the direction is set. The teams that win the next decade won’t be the ones with the most features. They’ll be the ones who learned, soonest, how to trust their machines.