How to Build High Agency (and Why Most People Don’t)

From whatishighagency.com · Updated May 2026

High agency is, by the consensus of the writers who study it most carefully, a learnable capability rather than a fixed personality trait. Shreyas Doshi: “some ppl are born/raised with high agency. It can also be developed later in life.”^[1] Cate Hall: “agency is something that can be deliberately cultivated by a lot more people.”^[2] Nir Eyal, in Harvard Business Review: “agency is not a personality trait. It is a learnable capability.”^[3]

The disagreement is not whether agency can be built but how. The short answer is that several specific mechanisms recur in the primary sources. The long answer follows.

The consensus that agency is learnable

For most of the twentieth century the dominant assumption in popular psychology was that traits like initiative, drive, or autonomy were largely innate. The recent literature on agency rejects this. Albert Bandura’s 1977 paper on self-efficacy was an early academic statement of the same point: expectations of personal efficacy are partly a product of experience and can be shifted by it. The modern startup-world version is more confident still.

Eyal’s HBR framing is the clearest mainstream version: agency operates through specific mechanisms in cognition and behavior, and each mechanism can be trained.^[3] Doshi’s thread treats agency as a hireable and developable trait,^[1] the central reason hiring on agency rather than talent is, in his account, a stronger long-term signal. Hall identifies three pathways: deliberate cultivation, exposure to high-agency people, and emergency mode — circumstances extreme enough to force the disposition into existence.^[2]

The takeaway from this consensus: if you wait until you feel high-agency to act high-agency, you will be waiting a long time. The trait is downstream of the action, not upstream.

Build, don’t only consume

Patrick Collison’s advice page contains the densest single statement on this: “make things. Operating in a space with a lot of uncertainty is a very different experience to learning something.”^[4] Reading about how startups work is information. Starting one, even a small one, is a different category of experience — the kind that produces the disposition.

The asymmetry matters. A person who has read fifty books about founders has fifty data points. A person who has shipped one small product has one data point of a different kind: they have evidence, in their own life, that the world responded to something they made. That evidence is what self-efficacy is built from. There is no shortcut that produces it without the construction.

The corollary: most reading-as-procrastination is the opposite of agency-building. Hall’s implicit critique of the tech self-improvement scene is that it tends toward consumption dressed as effort. The drill is to convert any idea you find interesting into something you can ship in days, not into a tab you keep open.

Choose your environment

Hall’s most concrete recommendation is the most often underweighted: “you can osmose agency from your environment if you’re exposed to the right kinds of people.”^[2] Patrick Collison says the same thing about the internet: “make friends over the internet with people who are great at things you’re interested in.”^[4]

The mechanism is straightforward and underrated: people calibrate their sense of what is possible against the people around them. If everyone in your daily orbit treats a given ambition as absurd, you will treat it as absurd. If you spend time, even online, with people who treat it as routine, the ambition starts to feel routine.

The single biggest lever most people have on their long-run agency is who they spend time with, and the second biggest is which conversations they choose to participate in publicly. Both are within reach.

Tighten the feedback loop

Patrick Collison again, in a transcript: “the main thing I think companies screw up at the pre-product-market-fit stage is sort of speed of iteration. Speed of fruitful iteration.”^[5] Drew Houston’s screencast on Hacker News is the same principle in personal-career form: when the standard channel for getting feedback is slow, build a faster one.^[6]

Two implications. First, the agentic move is not just to ship, but to shorten the time between shipping and learning. Second, fast feedback creates more chances to update, which compounds into the disposition. People who get many small signals per week feel more agentic than people who get one signal per quarter, even when the actions are identical.

Notice degrees of freedom

Hall’s definition of agency is partly diagnostic: “the capacity to both see and act on all of the degrees of freedom that life offers.”^[2] Most of the move is the seeing. The action, once you see clearly, is comparatively cheap.

The drill is small and daily: every day, identify one constraint you have been treating as fixed and ask whether it actually is. Some examples that recur in the literature: deadlines that aren’t enforced; required steps in a career path that nobody actually checks; introductions you’re assumed to need but could request directly. Most of life’s constraints are bundles, only one strand of which is load-bearing. Find that strand. It is often not real.

Reframe the timeline

A useful prompt that George Mack attributes to Peter Thiel: “how can you achieve your 10 year goal in 6 months?”^[8] The point of the question is not the answer. The point is the work of trying to answer it. That work forces you to identify which steps in the standard ten-year plan are real and which are convention.

Most often the answer is “I can’t” for some steps and “actually I could” for others. The gap between the two categories — what you must wait for and what you could collapse — is the most useful thing the question produces.

Stop waiting for permission

Stepfanie Tyler’s line is direct: “we don’t wait for permission that’s never coming.”^[10] Andrew Yeung makes a similar point: “we need fewer people who believe in the status quo and more people who are willing to be bold and take risks when everyone tells them ‘no’.”^[11]

The phrasing can sound aggressive and isn’t, quite. The underlying observation is mechanical: in most situations outside formal institutions, no single person has the authority to grant the permission you are waiting for. The structure is convention, not command. Recognising the difference is most of the work.

Naval Ravikant frames the same point from a hiring lens. The people he wants are “who just solve problems without even being asked to solve the problem.”^[12] The agentic move is to identify the problem before being assigned it.

Stay near the frontier

A specific 2026 addition that the older literature does not contain: stay close to the AI frontier. Tyler Cowen, summarised by Chris Barber: “the good news is you’re never far from the frontier, because we just created these things.”^[14] Sam Altman has framed the AI-era skill set similarly: high agency, idea generation, resilience, adaptability — all “learnable.”^[13]

The mechanism is direct: people who use the best AI tools weekly develop a faster intuition for what is possible than people who don’t. That intuition translates into agency, because agency operates upstream of capability — you act on what you can imagine being doable, and imagination is calibrated by tools.

Why most people don’t

If the mechanisms are this straightforward, why is low agency still “the default setting for most of us,”as Mack puts it?^[9] Three reasons recur in the sources.

The downside is local; the upside is delayed. The cost of acting agentically — looking foolish, being told no, having to redo work — is paid immediately and publicly. The benefits compound over years. Most people discount accordingly.

School is bad practice. Twelve to twenty years of being graded on doing exactly what was asked is, in aggregate, training for the opposite disposition. Hall: “okay, somebody just tell me what to do so I can be good.”^[2] The unlearning takes time.

Environments punish it asymmetrically. The social psychologist Katharine Greenaway’s critique of high-agency discourse is relevant here: the same action reads as agentic when one person does it and as troublemaking when another does. People who have been on the wrong side of that asymmetry have rational reasons to be more cautious about acting first and explaining later.

The cost of being wrong

A practical question worth answering before any of these drills: what does it cost you to push on a constraint that turns out to be real? In most cases, the answer is: very little. You look foolish for a moment, you re-route, you do the conventional thing instead. Five minutes of mild embarrassment is the typical worst case.

The cost of not pushing, on the other hand, compounds across thousands of small decisions over years. Treat agency as a probability bet running over a very long time horizon and the math works out unambiguously. The expensive failure mode is the one that doesn’t feel like a failure — the door you never tried.

References

Shreyas Doshi, X thread on agency vs. talent, 27 June 2020.
Cate Hall, “Can You Just Do Things?”, Asterisk Magazine, 2026.
Nir Eyal, “How Leaders Can Build a High-Agency Culture”, Harvard Business Review, March 2026.
Patrick Collison, Advice.
Patrick Collison, Running Your Startup transcript.
Drew Houston, On starting and scaling Dropbox (YC Library).
Eric Weinstein, Interview transcript, The Tim Ferriss Show #131, 13 January 2016.
George Mack, Thread on high agency thinking (X), 30 November 2018.
George Mack, High Agency in 30 Minutes, 2024.
Stepfanie Tyler, “High Agency People Are Annoying”, Bad Girl Media, October 2025.
Andrew Yeung, “What It Means to Be High Agency”, February 2024.
Naval Ravikant, On hiring high-agency people (Startup Archive), 2024.
Sam Altman (third-party clip), On high agency as an AI-era skill (X), February 2026.
Tyler Cowen via Chris Barber, On staying near the AI frontier (X), January 2025.
Paul Graham, “Relentlessly Resourceful”, March 2009.
Sahil Bloom, Reflections from a think-week retreat, May 2025.