Agency: What Philosophy, Markets, and AI Taught Me About Building Technology

I studied political philosophy because I was interested in power.

Who has it? How do institutions accumulate it? Why do people accept certain systems and reject others? And, maybe most importantly, what gives an individual the ability to act within those systems?

I didn't have the vocabulary for it at the time, but the question I kept returning to was agency.

I assumed I would spend my life thinking about that question through politics. Instead, I ended up exploring it through technology.

My career has moved from marketplaces, to home services, to crypto infrastructure, to artificial intelligence. On paper, those can look like disconnected industries. To me, they have increasingly felt like different versions of the same problem:

How do you give people more ability to act?

Learning distribution

My first real education in technology happened at Uber.

I wasn't an engineer or product manager. I was working directly with drivers: onboarding more than a thousand of them in San Francisco and eventually helping launch new ways of bringing drivers onto the platform.

What interested me wasn't only that Uber had built a better way to request a ride. It had built a system that allowed someone with a car to turn unused time and an underutilized asset into income.

Software had changed the range of actions available to a person.

That sounds obvious now. At the time, it changed how I thought about technology.

The most important products weren't necessarily those with the most sophisticated technology. They were products that took something previously difficult, inaccessible, or fragmented and made it possible for many more people.

At Ritual, I saw the same principle from another side of a marketplace. Instead of drivers, I worked with merchants. The technology mattered, but so did distribution: convincing businesses to adopt a new behavior, understanding why they didn't, and translating what I heard in the market back into something the company could use.

I began to understand that building technology and getting people to use it were not separate disciplines.

A product without distribution has limited power. Distribution without a useful product eventually collapses.

The interesting work happens between the two.

Learning that sales can change strategy

That became much clearer to me at Setter.

Setter started as a home-services and home-management company. My role was commercial. I sold home services, managed business development, and helped lead a team.

But the most important thing I learned wasn't how to hit a quota.

It was how a conversation with a customer could change what a company should become.

We began exploring a different model: instead of waiting for homeowners to encounter a problem and then selling them a service, what if home care could be packaged proactively and distributed through institutions that already had relationships with millions of homeowners?

Insurance companies immediately made sense.

They had aligned incentives. Homeowners wanted fewer emergencies. Insurers wanted fewer expensive claims. Setter could help both by making maintenance easier to understand and act on.

I helped open conversations with American Family Insurance and Nationwide.

Those conversations moved us upstream, from selling individual home projects to thinking about how home management could become a product distributed through much larger platforms.

Eventually, Setter was acquired by Thumbtack, which was itself expanding from a marketplace for individual home-service jobs into a broader home-management platform.

I wouldn't claim that one partnership caused an acquisition. Companies are more complicated than that.

But being inside Setter during that period taught me something I've carried into every company since:

Business development at its best isn't just selling what already exists. It can reveal what the company should build next.

The market talks back.

Good operators listen.

From accessing services to accessing technology

After Setter, I became increasingly interested in crypto.

Not because I thought every financial asset needed to move onto a blockchain. What fascinated me was the underlying permission structure.

Anyone could create a wallet. Anyone could interact with a protocol. Anyone could build on top of infrastructure that wasn't controlled by a single platform.

The philosophical question I'd been interested in at school had reappeared in a technical form.

Who gets permission to act?

At Underdog Protocol, my cofounder and I built infrastructure on Solana that made it easier for developers and companies to create and distribute digital assets.

At first, the problem seemed technical: blockchain infrastructure was difficult to use.

But the deeper problem was access.

Developers shouldn't need to understand every low-level detail of a blockchain to build something useful on top of it. A good abstraction increases agency by removing the complexity that prevents someone from acting.

We built APIs, distribution tools, and products that were used by hundreds of customers.

Some things worked. Many didn't.

The company went through the crypto downturn, changing customer behavior, new technical standards, and the normal existential uncertainty that comes with building a startup.

That period changed my understanding of entrepreneurship.

I had previously thought building a company was largely about conviction.

I came to understand that conviction without adaptation can become stubbornness.

You need a point of view strong enough to motivate people through uncertainty, but weakly held enough to admit when reality is telling you something different.

That tension eventually led us somewhere unexpected.

When software gained agency

In 2024, we started experimenting with AI agents.

The original idea was narrow: could a user describe what they wanted to accomplish financially and have software help them do it?

Then the model began expanding.

Instead of learning an interface, imagine telling software:

Find this asset.

Compare these opportunities.

Execute this trade.

Monitor this wallet.

Tell me when something changes.

Take this action when it does.

This was different from the software I'd spent my career working on.

For decades, most software had primarily helped humans read information or write information.

Agents introduced another primitive:

act.

We built Griffain around that idea.

It became a platform where users could interact with agents capable of performing actions across crypto and other digital systems. Tens of thousands of people used it. We built partnerships, products, revenue streams, and an ecosystem around the idea that software could move from answering questions to accomplishing objectives.

Eventually, my cofounder and I joined MoonPay, where I now work on bringing agentic products and financial infrastructure to a much larger platform.

And that experience has made the question of agency more complicated for me.

Agency creates responsibility

For most of my career, increasing agency felt unambiguously good.

Help someone earn money.

Help a homeowner solve a problem.

Help a developer build something without asking permission.

Help a consumer access a financial product.

Remove friction. Increase capability.

AI agents complicate that assumption.

What happens when the entity gaining agency isn't a person?

If software can interpret an objective, make decisions, move money, communicate with other systems, and act continuously on someone's behalf, then capability is no longer the only problem.

We also have to ask:

Who gave the agent authority?

What is it allowed to do?

How does a person understand what it has done?

Who bears responsibility when it is wrong?

What incentives shape the systems we delegate our decisions to?

How do institutions adapt when machines can participate in markets alongside people?

Those questions sound philosophical because they are.

They are also product questions. Business questions. Regulatory questions. Leadership questions.

I've come full circle.

The political philosophy student who was obsessed with systems and power became a salesperson, an operator, and eventually a founder. I spent years trying to make increasingly powerful technologies easier for people to use.

Now I'm interested in the harder problem.

What should we build once access is no longer the primary constraint?

Building institutions, not just products

Startups teach you speed.

When resources are scarce, you learn to move quickly, persuade people before you have perfect evidence, make decisions with incomplete information, and change direction when reality demands it.

I love that environment.

But the technologies I'm now interested in will require something beyond speed.

Financial systems, autonomous software, artificial intelligence, infrastructure: these technologies increasingly touch institutions that people depend on.

Building useful products inside those systems requires understanding not just how to start things, but how to make them endure.

How do you design an organization that can innovate without losing accountability?

How do you allocate capital when the consequences of being wrong become larger?

How do you lead people with different incentives, disciplines, and worldviews?

How do you build trust around technology whose capabilities are moving faster than the institutions governing it?

I don't think those questions have simple answers.

That's probably why I'm interested in them.

My career started with a fairly straightforward belief: technology can give people more choices.

I still believe that.

But I've learned that agency is not simply the removal of constraints.

Agency is the ability to act meaningfully within a system.

And every system, whether a market, company, government, or piece of software, contains rules, incentives, responsibilities, and forms of power.

The next generation of technology will give both people and machines extraordinary new abilities to act.

The question that interests me now isn't whether we can make that happen.

We almost certainly will.

The question is:

What kinds of institutions should we build around that power, and who do we need to become to lead them well?