What is agency?
I have returned to this question repeatedly, especially as a founder. Building a company asks you to commit your time, resources, attention, and a meaningful portion of your life to bringing something into the world. When the commitment is that large, direction becomes the ultimate question: What do I actually want to do? What do I want to see exist because I chose to spend my life on it?
It took me longer than I expected to ask those questions directly.
Three years ago, I began noticing that many of the beliefs I treated as defaults had histories. Some had formed around particular events. Others had entered through relationships, incentives, and environments that shaped me. I had acted from these beliefs without recognizing them as beliefs. They simply felt like reality.
That realization made the question of agency more difficult. I could make decisions, take risks, and work relentlessly. Yet how much agency did I have if I had never examined the assumptions deciding what seemed worth doing?
I now think agency has at least two layers. The visible layer is action agency: the capacity to turn an intention into reality. Beneath it is belief agency: the capacity to surface, test, and revise the beliefs from which intentions arise.
Belief agency comes first. Action can take us far, while beliefs determine which directions appear possible, valuable, or even visible.
Some beliefs hold the rest of the stack in place
Your beliefs form a stack. Surface beliefs rest on deeper ones. “I should work in finance” sits on “money equals success” which sits on “external validation matters.” Most people only see the top of the stack.
Among the strongest founders I know, speed is the visible trait. A less visible one is their willingness to inspect assumptions that everyone else treats as fixed. They notice when a belief has stopped explaining reality, even when that belief once contributed to their success.
To have belief agency, you need to:
-
Make your beliefs explicit. You can’t change what you can’t see. Write them down. Most people have never done this.
-
Identify the load-bearing beliefs. Some beliefs are decorative. Others are structural. Change a load-bearing belief and your entire worldview shifts. “The future is predetermined” vs “I create my own future” – these aren’t just different opinions. They’re different operating systems.
-
Test through action. Beliefs aren’t abstract. They predict outcomes. If your beliefs consistently lead to poor predictions, you’re using the wrong model. Update your priors.
Hidden assumptions set the boundaries of action
The hardest beliefs to inspect are often those that no longer feel like beliefs. Experience, incentives, and environment can reinforce an assumption until it appears to be a simple description of reality.
I see this constantly in Silicon Valley. Smart people with all the action agency in the world can code, fundraise, and ship. Their decisions can still rest on assumptions they have never articulated: what users want, what growth must look like, which risks are acceptable, or what kinds of company can exist.
Competence can make these assumptions harder to notice because they produce coherent action. Success can reinforce them further. The person becomes more effective inside the model without learning whether the model itself is sound.
A person rarely attempts an option their model has already ruled out. Hidden assumptions therefore define the boundaries of action before effort begins.
Visible beliefs can be tested against reality and values
A belief can be revised only after it can be stated.
Start by noticing when you say “that’s just the way it is.” That’s code for “I’ve never questioned this belief.” Question it. Where did it come from? What evidence supports it? What would happen if the opposite were true?
Visibility creates a second question: which beliefs deserve to stay? Restacking without an objective is arbitrary. Every belief stack already serves some metric, even when that metric has never been named. The metric may be security, autonomy, belonging, service, craft, care, or creation. Values establish direction. Beliefs model which actions will move us toward it.
Consider the belief that money equals success. It may support security, expand autonomy in some contexts, or narrow a life around external validation. Examining the value beneath the belief can reveal that the model needs revision, or that the objective itself was inherited.
This gives a belief two tests. Does it remain accountable to evidence? Does acting through it advance values the person has actually chosen? A factual claim does not become sound because it is motivational. A working belief about the self, such as “I am capable of learning anything,” can be tried provisionally and revised through action. Usefulness creates a reason to experiment. Reality still gets a vote.
Other people make the contingency of our own models easier to see. The conservative banker and the anarchist artist can organize coherent lives around very different premises. Their existence does not prove either model correct. It reveals that what felt inevitable may have been one workable arrangement among several.
Experiments can then generate evidence. Hold a working belief lightly for 30 days and act as if it were true. “I’m creative.” “People want to help me.” Observe what changes in your attention, behavior, and outcomes. The result may strengthen the belief, qualify it, or expose where it fails.
Before asking which belief is useful, ask what the belief is useful for.
Agency is alignment from values to action
A belief draws its power from its position in two interacting structures:
Beliefs → perception → action → outcomes → belief revision
Values → what counts as a worthwhile outcome
The first structure is a feedback loop. An outcome returns as evidence that can update the belief or reveal a mistaken interpretation. The second gives the loop direction. It explains why one outcome matters more than another. The two interact, and no single arrow fully determines the next.
Consider a founder deciding whether to continue a product after adoption falls below plan. One belief says that a good product should pull users immediately, so slow growth looks like rejection. Another says that a new behavior may begin with a small group of deeply engaged users, so the same result directs attention toward retention. The observed numbers are identical. The beliefs change what each founder notices, which explanation seems plausible, and what each one does next.
Their actions then change the evidence available to them. The first founder may shut the product down. The second may interview retained users and run another experiment. Each action produces a different outcome, and that outcome becomes the material from which the original belief is judged. The belief has helped produce the evidence that will later confirm or challenge it.
Values shape the loop as well. A founder optimizing for rapid scale may interpret the evidence differently from one trying to develop a durable craft or serve a narrow community. The value influences which outcomes count as progress without dictating the decision.
This is why alignment requires continuing work. A load-bearing belief changes what a person sees, attempts, persists through, and learns from. Outcomes test the model. Values determine why the test matters. Reflection reconnects them before the next action.
Action agency carries an intention into the world. Belief agency brings experience back into the model from which the next intention will arise.
Ultimate agency offers neither control over reality nor permission to believe anything. It is the capacity to keep one’s model of reality, chosen values, and actions in active alignment.
What am I trying to protect, create, or become, and which beliefs help me see reality clearly enough to act toward it?