I think that in tech we need a tad more reason. We romanticise things a tad too much, and we categorise people way too quickly. Especially when the word is .
For me, agency is directing your choices and acting toward goals you have chosen. Not performing someone else's ambition. It is not about "quitting your job and starting a company"; that is definitely not the definition of agency.
Silicon Valley takes words like "taste" and makes them unusable to me. It also takes things poor people have been doing forever, rebrands them, and makes them socially exciting:
- Immigrant speaking several languages → polyglot
- Immigrant → expat
- Hustle →
- Gossip → backchanneling
What do you actually want?
I placed multiple people in 2025 while working at an early stage startup. I exercised a ton of agency during my jobs. You can run a business while having a job, btw, depending on your work authorisation and commitments.
You do not need to want to raise venture capital. There are so many incredible entrepreneurs that never raised a dollar. You do not need to want to start a business. There are incredible operators who changed the course of history by doing their life's work in the right place at the right time. You do not have to be in tech to be happy, nor in finance, nor in law or medicine. You are not forced to want anything.
You want to become a doctor, but make no effort to get into pre-med? You want a job in AI or an early stage startup, but don't use AI tools or read about tech? Do you really want it?
If eating chocolate makes me happy, I will eat chocolate. If tech interests me, I will code and do gimmicks in my free time and learn about the industry. I don't need to be "managed" or "motivated" into either; I do it because it makes me happy.
It is not because I am "high agency"; I am a human maximising my happiness function. A lot of people are mistaken for having low agency when they are doing something they think they should be doing. Every human has agency, but might not have clarity on what to exercise it on.
What makes you hungry?
How do you develop agency? Honestly, this is an odd question for me. You either are fed, or you are not; if you are not, you either are hungry, or you are not, and that is ok.
My motivation to never stop can't be "learned" or "replicated." I want to retire my mom. I am first-gen, grew up as an immigrant raised by her and my grandma. My grandma never went to school, nor learned to read or write. You can't "develop" this situation for yourself.
Because of this, I value education like a priceless thing. It is a privilege to be able to read, and write, and to learn something new every day. I love what I do because I understand what it means not to have it. Of course I listen to podcasts, read the news and grok new models. It isn't a burden; it is something I get to do.
I honestly feel like I live in a renaissance: from playing the piano and speaking 6 languages to playing with sci-fi AI, meeting founders and investors, and helping in the small ways I can. You wouldn't be able to pay me to stop doing this.
But this is me. Dig inside yourself: what makes you feel like "there's no way I am paid to do this"? Find a way to do it as much as possible.
A happiness maximisation function
In an era where intelligence is abundant and a commodity, we get to ask: what makes me happy? I am long humans. Figure out your , and I think you will naturally want to maximise it.
What does mine look like? Honestly, it evolved over the years, and I hope it will evolve again. I mean maximising happiness and minimising pain given my constraints, not measuring happiness perfectly.
Let's start with a : a rule assigning one output to each input. Here, x is our . Take the function below:
If this is the function we want to , it is our . Its maximum is 9. The input that achieves it is 3: the maximum tells us the value, while the tells us which input achieves it.
Now add a : 0 ≤ x ≤ 2. We can no longer choose 3. The best allowed input is now 2, giving an output of 8. All inputs satisfying the constraints form the .
Each input x has one output H(x). The height of the point is that output.
With two decision variables, x and y, each input is a point in a plane. The feasible set is now a region. The next example maximises a function of both variables, subject to lower bounds, upper bounds, and a circular constraint.
In the graph, DV1 is x and DV2 is y. The shaded area is the feasible region. Each dashed has a constant objective value, x + 2y = c. The arrow indicates increasing values, and the dot marks the optimum. An holds at equality at that point; an is satisfied strictly. For g(x, y) ≤ b, the slack is b − g(x, y): zero when active, positive when inactive.
Agency as freedom
Agency is how we exercise our freedom: figuring out what we want and acting towards it. But if my happiness maximisation function has constraints, I need to understand which freedoms I actually have.
Freedom is not the same as ability
Think of someone in a wheelchair who cannot stand. Technically, no one is preventing them from getting up; if they did get up, no one would be pushing them back into the chair. But do they actually have the ability to do it? No. That is a different question.
Now ask: do I have the resources, tools, and ability to act? Being left alone does not supply them. concerns the real opportunity to act.
Before calling someone lazy or "low agency" for not doing xyz, ask: do they actually want to? If they do, are they unable to, or unwilling to make the effort? Those are different questions.
Ugly truth: people have different constraints. Someone on a visa may be able to apply to but unable to quit their job and join the community if they get in. Someone providing for their kids, wife, husband, or parents may need to stay employed longer before starting a business. That is not laziness or a lack of ambition.
Before you tell someone to drop out
I literally had a Stanford student call me this summer asking whether they should drop out of school and start a business, because everyone around them was doing it. They had been sold the idea that looking "high agency" meant quitting everything and risking it all.
I told them: I do not know you. You are a stranger to me, and I cannot give you direction on your life, nor financial or legal advice. But we can unpack your happiness maximisation function, starting with the constraints.
Were they on a visa? Yes, . F-1 status comes with enrolment requirements. They could not assume they could stop studying and remain here to build a company. Exceptions and other immigration paths exist, but they needed to understand their own situation.
They were from a country in Europe, where starting a company was another option. But they wanted to build it here. They had not considered what dropping out would mean for that.
I asked what their parents did and whether they could support them financially. No. Monthly living costs? 2k. A plan to raise? They had not thought about that either.
Letting other people dictate how to use your agency is a terrible way to exercise it.
Figure out what you want, then be honest about your constraints. A more complicated equation is still yours to work through as an adult.
Price the risk
Risk taking is a function of , and some risks are plain dumb. I started my career as a trader. At JPMorgan, Goldman Sachs, or Bank of America, your job involves . If your boss asks why you made a trade, you need facts, sources, and back-testing backing your opinion. "Because I feel like it" or "because others are doing it" is not an answer you get away with from a trading seat.
So if there are entire careers based on pricing risk, why shouldn't we apply this same thought process to our lives? is how much uncertainty you are willing to accept. is how much downside you can absorb. They are not the same.
Before getting my green card, casually quitting a job to "figure out what to do next" meant putting my ability to stay in the country at risk. After getting it, my right to stay no longer depended on that job.
Does this mean I was lazy before, or not entrepreneurial enough, or not willing to take risks? No. It means I wasn't stupid enough to take a stupid risk. My ambition never changed, but suddenly the choices available to me did.
My function, right now
The visa constraint is gone. The monetary one is not. NYC is expensive, and I am not a nepo baby with assets or investments paying for my life. I can't work for free. I want to provide for my loved ones and retire my mom.
My joy comes from helping people succeed, learning and building, making beautiful things, reading, writing, surfing, running, working out with my husband, and kissing my cat even when she doesn't want to be touched.
I joined SPC to figure out my life's work. But everyone is an artist until rent is due, and exploration doesn't pay rent. Money does. I saved in my previous job to fund this time; I still have only 24 hours in a day and 7 days in a week.
Having a green card and savings lets me quit a job, stay in this country, and explore at SPC. I worked hard for this, but it is still a privilege, for which I am so grateful every day. Why am I bringing this up? To explain to you that: this isn't just something I could do on a whim, at any time, in any situation.
Sama has a different equation
I hope one day money will no longer be an issue, and I can ask: "What is the highest leverage thing I can do with my life? What is the most important problem I can solve?" Sam Altman's existing wealth gave him the breathing room to take the opportunity to do his life's work at OpenAI. In December 2024, he reportedly earned about $76,000 a year and held no in OpenAI.
From a trading seat perspective, given my happiness optimisation function, taking a job with a small salary and 0 equity is dumb for me right now. Would I love to do something like this one day? Oh, I would love to. Can I afford to, right now? No. I have the permission, not the practical freedom. I have the negative freedom, not the positive freedom for this.
Other folks as rich as Sam, or richer, prefer chilling on a beach. There is nothing wrong with that. If they claim they want to do what Sam does but take no action, yes: these people then are low agency. If they want something else and act towards it, then these people are high agency. You exercise agency, your free will, towards what you actually want + can.
I adore Sama, as I adore so many folks who inspire me. But admiring someone else's life does not make it the right life for everyone. People are non-linear and polynomial equations, not a binary label.
Explore, learn, act
Can I explore at forever? No. At a certain point, after exploring, you need to exploit.
In , an observes its environment, takes an action, and receives a and a new state. Its determines how it chooses actions. The objective is to maximise expected return: rewards accumulated over time.
The agent does not initially know which actions give the highest return. means taking actions to learn about their outcomes. means choosing what currently looks best.
One way to choose between them is . With epsilon set to 0.2, each decision has a 20% probability of choosing a random action and an 80% probability of choosing the highest-valued action according to its current estimates. Random choice can still select that same action.
For a period of time, it is ok to spend 90% of your time exploring and 10% exploiting. You are discovering the surface area that exists out there, and even if you explore 90% of the time, you might still not discover everything. But at a certain point, with a finite runway, you need to exploit more: reverse that to 80-90% exploitation and 10-20% exploration. Otherwise, you might end up with little to nothing.
Initial value estimates: A = 4, B = 6, C = 3. The greedy action is B.
RL equations
How does the estimate change after observing a reward? In , Q(s, a) estimates the return from taking action a in state s and then acting optimally. After observing reward r and next state s′, the update uses a target y: the reward plus the discounted highest estimated value in that next state.
Here, the one-state uses γ = 0 and α = 0.25, so the target is just the immediate reward. Starting from an estimate of 6 and observing a reward of 8 gives:
In broader RL, actions can change the next state, and their values include later rewards. Deep RL uses neural networks to approximate functions such as values or policies. Epsilon-greedy is one exploration strategy: with sparse rewards, random actions may rarely find useful feedback. Other methods guide exploration using uncertainty, visit counts, novelty, or learning progress, although novelty alone need not improve the task reward.1, 2
"Context is not static." A uses the outcome of an action to inform the next one. That is the point of Everything Should Be a Closed Feedback Loop, and it applies to the equation itself: what you want and what you can do may both change.
A person is not a single score
We need to stop categorising people as having "agency" or "no agency". There are a lot more axes to look at when reading a person; it is a multidimensional problem. People are non-linear and polynomial equations, not a linear step function.
And reducing excellence to a few chosen variables, is lazy at best and a disrespect to said excellence.
From Non-Linear Polynomial Lives
The 3D view could not load.
Mentorship, and doing it scared
The best companies don't have managers, but mentors. When someone answers "How are you?" with "I'm managing," they're telling you there's a problem. As a business owner, you shouldn't pay for problems but for solutions. If you need to manage an employee, you are bad at hiring: you hired a problem, not a solution.
In this new world of AI, anything is possible and anything can be built. It's beautiful. So, if someone does not do something, there are very few options:
- They can't, or they are unable. This is no longer the case. Anything can be done.
- They are able, but they are lazy or unwilling. Then this means it is your fault: you hired the wrong person.
- They are able and willing. Usually, they didn't have the context, the environment, the knowledge, tools, or network needed to make something happen. This person does not need to be managed. This person needs to be mentored.
High agency people do not need to be managed, but mentored. This person wants and desperately needs mentorship: human connection.
Frankly, there might not even be lazy people, but people working on the wrong thing, at the wrong time, in the wrong place. It is an inefficiency in the market. What if everyone could work on what they actually loved? This becomes a resource allocation problem, like a capital allocation problem. And yes, it becomes not only a hiring problem, but also an inner search that everyone should do within themselves.
Doing it scared
I have the freedom to post YouTube videos, and to post on Instagram and LinkedIn, yet I don't do it that much. Why? Honestly, it's scary. I don't see agency as "do it all immediately"; it's a mix of drive and willingness to do it scared.
When I posted my first post on LinkedIn, I remember sweating cold. I was terrified. But I want to get better at making videos and posting online. In a way, agency is hustle, and being willing to climb cringe mountain and do it scared.
Write your own equation
Being honest about your constraints does not mean you are lazy. Your equation is just a tad more complicated, but it is still your job as an adult to write it out for yourself.
If you are in your 20s, I invite you to read this, this, and this.
And honestly, there is nothing more low agency than forcing yourself to do something and lying to yourself that you like something just because you feel like you need to like it. Choosing not to want what everyone around you wants is also exercising your freedom and agency. And once you have honestly accounted for what is getting in the way, if you are not doing something, maybe you just don't want it badly enough, and that is ok. High agency is figuring out what makes you happy, and not inventing excuses for not taking any action, even if a baby step, towards it.