Authors: Alana Levin, Kayvon Tehranian

Translation: TechFlow

TechFlow Introduction: Variant has just completed raising $222 million for its fourth fund, shifting its investment theme from "ownership economy" to "autonomy." Partners Alana Levin and Foundation founder Kayvon Tehranian jointly released ten investment hypotheses, covering directions such as AI Agents, computing power, open source, and multi-model systems. This list essentially serves as a public scorecard — they laid their bets on the table and waited for the market to score.

Variant invests in technologies that expand autonomy — allowing individuals or organizations greater freedom to build, customize, and act in their own way.

Automation helps developers, individuals, and organizations do more with fewer resources. Well-executed automation provides users with new capabilities while giving them greater control over their technology stack.

This scope is vast, so during our exploration, we anchored our thinking direction with a set of hypotheses.

Here are ten working hypotheses formed during discussions over the past few months. They help us decide where to invest time and attention. Some hypotheses may be incomplete, and some may be wrong. But this exercise forces us to clarify our implicit assumptions, providing us with a framework to judge: how the world will change, what products matter, where value will concentrate, and who can capture it.

Agents will become the main source of internet traffic

Today’s software defaults to a human clicking buttons and viewing interfaces. However, an increasing number of products need to be designed for both humans and Agents simultaneously. Agents operate around the clock, can be infinitely replicated, and have behavior patterns and incentive mechanisms vastly different from humans.

Compute shortages will last quite a long time

Scarcity makes infrastructure valuable. Startups that can reduce compute costs, increase utilization, or open new sources of supply will become increasingly valuable.

The role of open source in the AI technology stack will become more significant

For enterprises, open source offers a viable path to compete with cutting-edge labs. For developers and consumers, it lowers the cost of experimentation and broadens access to cutting-edge models.

The barrier to using cutting-edge AI will continue to rise

Cutting-edge AI is becoming increasingly "permissioned." The demand for open, unlicensed alternatives will grow as developers need to retain autonomy.

The number of truly important models will grow by an order of magnitude

As models become more specialized, the ability to select and orchestrate models will be more valuable than building around a single model.

Multi-model systems will become the default choice

Applications will manage the complexity of distributing tasks across multiple models, without users needing to worry about these aspects.

Harness (orchestration layer) is the core of value accumulation in the AI technology stack

Models are becoming easier to replace, and enduring relationships shift to orchestrated models and products that provide user experiences. Here, Harness refers to the application layer that wraps the models — it is responsible for scheduling, routing, and context management, and is the layer that directly faces the users.

Current Agent systems are mostly single-mode; the frontier is multi-person collaboration

The challenge shifts from "building a single Agent" to "coordinating shared contexts, permissions, workflows, and collaboration across organizations." Defining and owning the control plane is an important leverage point.

AI will create low-cost alternatives to many expensive services in society

Sectors like healthcare, education, and legal services are particularly well-suited for newcomers to rethink the economic model of knowledge delivery. Some of the most innovative solutions may come from directly empowering end users.

AI will make many previously impossible markets economically viable

As Agents can automate much of the work required to found and expand companies, entrepreneurship in more fields becomes viable, allowing founders to focus on the most challenging problems.

We are releasing this list as a public scorecard at the beginning of Q3 2026. We look forward to tracking which of these judgments withstand the test of time and will continue to update as our understanding evolves.

If you are building products around these hypotheses, we would like to talk. Especially if you disagree with some of these judgments — let us know where we're wrong.