**In this essay, I argue that agency is a continuum—one that, over time, steadily shifts execution from human hands to external tools, fundamentally transforming creativity, skill, and value. I conclude with what we must do. **

Changes in expression of human creativity can be mapped alongside the shifting boundary between intent and execution. Over centuries, we’ve built tools—agents—that increasingly absorb the technical work of creation. This evolution forms what we might call the agency continuum: a spectrum of tools, practices, and technologies that act as external agents, executing our ideas with increasing sophistication.

Two domains show this continuum with some clarity: image-making and software development. Though they evolved in different historical contexts and serve different purposes, both reveal the same underlying pattern: a movement from low-tech, manually intensive agency to high-tech, automated, even generative agency. Along this continuum, the nature of joy changes, the sources of value shift, and the definition of human creativity transforms.

For most of human history, creating an image meant painting—a craft requiring years of training—understanding light, shadow, perspective, anatomy, colour, and composition. Intent and execution were fused: you imagined, and you physically rendered. Even here, we see examples of great artists employing “external (to them) agency” — apprentices, assistants or even a whole atelier to complete large pieces of work. Idea still belonged to the artist but was executed with help. Any “computational” changes was product of the mind and skill of the hand - for e.g. a painter portraying a monarch or a patron might bring about changes which might be different from reality - making some features prominent etc.

Then photography arrived and split the two. The camera became an agent that executed the technical details—exposure, focus, chemistry—while we concentrated on framing and meaning. Human creativity shifted toward choosing, framing, noticing—a different type of artistry. As technology improved, photography became increasingly automatic: eventually digital, then computational in its true sense. Today, smartphones perform complex processing behind the scenes, turning a press of a button into a polished image.

Now AI image generation completes the circle. Like painting, it enables pure imagination—but without technical barriers. You describe an idea; the agent renders it. Execution is instant, abundant, and often superhuman (in the narrowest sense of skills required vs output). The user becomes the originator of thought; the agent becomes the executor of form.

Software development followed the same path. Early programming resembled painting (in a sense of skill required): humans wrote punch cards and assembly, directly manipulating the machine. Again, intent and execution were inseparable.

Compilers and high-level languages introduced the first major break. They automated memory management, hardware interactions, and optimisation. Frameworks and cloud services pushed agency further, swallowing deployment, scaling, and infrastructure.

Today, AI-assisted coding represents the equivalent of generative imagery. Developers specify high-level intent—architecture, constraints, behaviour—and agents write, refactor, and debug the code. Execution has become automated and abundant.

The programmer describes intent. The agent produces execution.

The separation of idea and implementation becomes complete.

Painting has a joy rooted in craft: the tactile struggle, the slow mastery of technique, the physical intimacy between hand and material. Photography has a joy rooted in noticing the world. AI image generation has a joy rooted in instant imagination and quick iterations. Similarly, low-level programming offers the joy of control and mastery, high-level programming the joy of problem-solving, and AI-assisted programming the joy of creation through thought.

Across the continuum, the activity changes—but the joy endures. Humans find delight in expression, insight, and making, regardless of whether the tool is a brush, a camera, a compiler, or an AI model.

1. Value Tracks Scarcity, Not Execution

Scarcity—not difficulty—best predicts value.

The most expensive art ever sold is Salvator Mundi which is valued for because it is inextricably tied to the artist’s hand and intent, rather than simply its execution. Even the most perfectly executed, technically impeccable painting may pale in front of brilliant intent.

The most expensive photograph ever sold is Le Violon d’Ingres which is valued for it’s intent rather than technical execution. Even so, it was 40x cheaper than the most expensive painting sold—not because photography is less meaningful, but because it is less scarce.

AI imagery, infinitely reproducible, drops scarcity almost to zero. Its value lies in context, authorship, and application—not the artifact itself.

Value follows scarcity. Scarcity shifts with agency.

2. The Value Curve Flattens as Agency Increases

On the left side of the continuum (painting, assembly language), value is concentrated among a tiny elite.

As agency grows, something happens: value becomes more evenly distributed.

Photography created entirely new industries - newspaper photojournalism, advertising photography, fashion shoots, wedding photography etc.

These markets did not exist before the camera. Photography did not replace painting—it expanded the total creative economy.

Demand exploded and participation broadened. Value spread across far more practitioners.

For an average person, it is easier to make money out of photography than by painting. There are 5 times more professional photographers than professional painters in the USA (as derived from this data).

The same is happening in software: Early programmers were a tiny priesthood. Then, High-level languages created millions of developers. Now, AI coding assistants will enable even more people—designers, analysts, entrepreneurs—to create software without traditional expertise.

Higher agency → more applications → larger markets → more earners.

Execution becomes standardised. Application becomes the economic frontier.

3. Barriers and Gatekeepers Change, But Never Disappear

Even as execution becomes easier, scarcity persists through new gatekeepers.

In photography:

equipment is expensive

a few brands dominate (Canon, Nikon, Sony)

lenses and software lock people into ecosystems

distribution once depended on magazines, agencies, galleries

Agency democratised execution but introduced industrial gatekeepers.

In software:

cloud platforms (AWS, GCP, Azure) control computational resources

frameworks and toolchains create structural dependencies

AI models are trained and run by a handful of organisations

Higher agency reduces expertise barriers but increases infrastructure gatekeeping.

Scarcity moves from skill → access → platform control.

How We Adapted Before

Every technological leap clarified rather than erased the human role:

Painting survived photography by becoming more expressive and conceptual.

Photography survived digital automation by doubling down on vision and narrative.

Low-level programming survived abstractions by anchoring the foundations of computing.

At every stage, humanity adapted by moving up the stack—toward meaning, perspective, and intent.

What We Should Do Now

As AI becomes a high-agency collaborator:

Elevate intent as the core creative skill. 1.

Cultivate taste, judgment, clarity, and conceptual depth—the new scarcity. 1.

Preserve artisanal forms of creation—they gain value, not lose it. 1.

Educate for thinking, not just doing. 1.

Find newer applications and markets for your ideas.

Execution is no longer scarce. Ideas and applications are.

No posts