AI Builders Are the New Starving Artists

The workbench is only half workshop. The other half is warehouse, storefront, and media desk. Illustration generated for this essay.

The hour after the thing works

At 1:47 a.m., the prototype finally does what it was supposed to do.

The laptop throws cold light across a desk crowded with coffee rings and charging cables. A terminal window stops spitting red errors. The button works. The workflow completes. For a few clean seconds, the room feels enormous.

Then the builder opens a blank tab and types: How do I explain this?

That is the hour in which an independent AI builder begins to resemble a starving artist.

Not because every agent, wrapper, automation, or generated interface is art. Most software is useful infrastructure. Some of it is dull on purpose. A clever accounting workflow does not become a painting because its creator used imagination.

The stronger comparison is about position. You have an idea. You acquire enough craft to make it visible. You spend an unreasonable number of nights on something nobody commissioned because the possibility will not leave you alone. Then you carry it into public and discover that making the thing was only one job.

Now you must make people care.

This is not a claim that all software is art

Calling every act of creation “art” drains the word until it can hold almost anything. The interesting claim is narrower:

AI builders are inheriting the incentives, anxieties, and labor conditions of independent artists, whether or not the objects they make qualify as art.

The resemblance appears after creation.

Independent artist	Independent AI builder
Develops a voice and body of work	Develops a point of view and product language
Pays for materials, rehearsal, or production	Pays for models, compute, APIs, hosting, and tooling
Competes for a gallery wall, playlist, stage, or feed	Competes for a demo slot, directory, app store, feed, or search result
Translates difficult work into an artist statement	Translates technical work into a landing page and demo
Balances commissions with personal practice	Balances client work or a job with independent building
Learns that visibility and quality are related, but not identical	Learns exactly the same thing

At the desk, these categories feel less tidy. The builder may be designer, engineer, researcher, customer support, bookkeeper, video editor, and sales department before lunch. The musician learned this arrangement earlier: write the song, record it, cut the clip, post it, answer comments, pitch playlists, sell tickets, repeat.

The independent builder is now holding the same stack of hats.

AI lowers one wall and reveals the next one

Generative tools can reduce the distance between an idea and a working prototype. That does not mean they reduce it evenly, or that they always make experienced engineers faster.

The evidence is messier than the sales copy. A 2026 meta-analysis of 23 studies found a moderate positive productivity effect from generative coding assistants, with substantial variation across settings. In a randomized study of experienced open-source developers working inside mature repositories, early-2025 AI tools increased completion time by 19 percent. Different people, different tasks, different codebases.

Still, for many bounded projects, the wall is lower. A person who once needed a full team can now sketch an interface, generate scaffolding, connect an API, produce copy, and assemble a credible demonstration from the same chair.

That changes the bottleneck.

When fewer hours are required to produce the first version, more versions enter the world. The scarcity moves downstream—from execution toward selection, trust, attention, and sustained use.

flowchart LR
    A["Imagination"] --> B["Prototype"]
    B --> C["Working product"]
    C --> D["Explanation"]
    D --> E["Discovery"]
    E --> F["Trust"]
    F --> G["Use or purchase"]
    G --> H["Support and survival"]

    I["AI reduces some production cost"] -.-> B
    I -.-> C
    J["Attention remains scarce"] -.-> D
    J -.-> E
    J -.-> F

The cheap prototype is not the finished economic object. It is the beginning of a longer negotiation with strangers.

The second product is the explanation

Imagine two browser tabs.

In the first, the system is alive. It can ingest documents, maintain state, invoke tools, and recover from an interrupted task.

In the second, the builder has written: “Your intelligent workspace for the future of work.”

The first tab may contain months of thought. The second sounds like twelve other products posted that morning.

This is not merely a copywriting failure. New software often has no stable category. If a builder explains the mechanism accurately, the audience may not understand why it matters. If the builder borrows a familiar label, the product becomes legible but interchangeable. If they promise a revolution, they join a feed already shouting.

So the explanation becomes a second act of design. What must a stranger see in thirty seconds? Which complexity can remain hidden? What evidence earns another minute? Which sentence distinguishes a real capability from a staged demo?

Artists know this discomfort. The work is meant to speak, but the room keeps asking for a statement.

Everyone becomes a media company, badly

At some point, the code editor closes and the camera opens.

The builder records the screen. The cursor jitters. A neighbor’s vacuum starts behind the wall. The first take is too slow; the second crops the important click; the fifth has energy but skips the explanation. Two hours disappear into a forty-second clip.

This is the hidden transfer of labor in the new creation economy. Production tools improve, so public distribution becomes everybody’s private responsibility.

LinkedIn makes the shift especially visible because professional identity and content performance occupy the same feed. The platform says its ranking systems use signals including views, reactions, comments, shares, dwell time, relationships, freshness, and subject matter to order content (LinkedIn Help). LinkedIn’s own engineering team also reports that daily content volume has grown, increasing competition for attention (LinkedIn Engineering).

The predictable adaptation is not mysterious. People compress work into a recognizable performance of work. The quiet six-month repair becomes “five lessons from shipping.” The failed experiment becomes a carousel. The ordinary career—a useful job done carefully, with no claim to disruption—becomes strangely difficult to post.

The problem is not that builders communicate. Clear public explanation is part of responsible work. The problem begins when the shape favored by distribution starts determining what gets built, or when the performance of momentum replaces the slower evidence of progress.

Abundance does not abolish gatekeepers

The old gatekeeper stood at a door: a publisher, label, broadcaster, investor, procurement team, or software distributor.

The new gate is a field of rankings.

Search position. Feed distribution. App-store placement. Product-directory votes. Newsletter opens. Follower graphs. Model marketplaces. Recommendation systems. The door appears open because anybody can upload. Past the threshold, visibility is still allocated.

Music offers a useful, imperfect comparison. Spotify reports that more independent artists are earning meaningful royalties, while its own figures also show how far down the ranking one travels before annual Spotify income reaches modest levels: in 2025, the platform says the 100,000th-highest-earning artist generated more than $7,300 in Spotify royalties (Spotify, Loud & Clear 2026). Those are company-reported figures and royalties are not the same as take-home income. They still make one tension concrete: access to distribution can widen at the same time that attention remains steeply concentrated.

AI software may follow a similar curve. More people can make. More people can publish. More people can find a small audience. None of those facts guarantees a living.

The starving part is not romantic

There is a dangerous glow around the phrase “starving artist.” It turns unpaid labor into character and financial precarity into proof of seriousness.

At the actual desk, romantic deprivation looks like a declined card for an API bill. It looks like support messages answered during a lunch break. It looks like postponing a dentist appointment while a product with 800 registered users produces $73 that month.

The analogy should not celebrate that arrangement. It should expose it.

Independent creators are often told that passion is a business model in waiting. Keep posting. Build in public. Find your audience. Be consistent. The advice contains useful actions, but it quietly transfers platform risk, market research, sales, and promotion onto the person making the work.

A builder can love the product and still need rent. They can refuse growth-at-any-cost and still want customers. They can seek recognition without pretending money is vulgar. Appreciation does not decide whether the work is art. It decides whether the maker receives recognition. Payment decides whether they can keep making under the same conditions.

What becomes valuable when making gets cheaper?

If production costs fall, value does not disappear. It rearranges.

Four things become more important:

1. Judgment. Knowing what deserves to exist, which trade-off matters, and where generated output is wrong.
2. Originality. Not novelty for its own sake, but a point of view formed from experience that a generic prompt cannot supply.
3. Trust. Evidence that the system works, protects data, survives failure, and will still be supported next month.
4. Distribution. A durable route to the people for whom the work is genuinely useful.

This is why the future may contain far more software and fewer stable software products. A capability might be composed on demand, bundled into a larger service, or given away to attract an audience. In that world, craft still matters. So do reliability and maintenance. But the standalone artifact competes with an ocean of things that are good enough to try once.

The builder’s defensible advantage may be the strange thing they noticed, the community that trusts them, the hard-won workflow they understand, or the discipline to keep repairing a tool after the launch thread disappears.

I want collaborators, not spectators

The creator economy trains us to count spectators: impressions, likes, followers, wait-list names. Those numbers can help, but they are thin substitutes for participation.

For independent AI work, the more useful question is often not “How do I make this go viral?” It is “Who sees enough of the problem to help make the next version truer?”

That person might be a designer who can make a difficult workflow legible, a domain expert who can identify the dangerous assumption, an engineer who cares about failure recovery, a researcher who can design a real evaluation, or an industry partner willing to test the system where the constraints are not imaginary.

This essay is an invitation to those people. I am interested in collaborations where the artifact is allowed to remain specific, the claims can be challenged in public, and distribution follows evidence rather than replacing it.

At 2:18 a.m., the working prototype still sits on the desk. The empty launch post sits beside it.

The goal is not to turn every builder into an influencer. It is to build better paths between careful work and the people who can use, test, fund, or improve it—so the maker can return to the workbench before the light comes through the window.

Sources and limits

• The artist comparison is an argument about incentives and labor, not a classification of software as art.
• Productivity results for AI coding tools are sensitive to task, expertise, codebase, and study design. The two studies cited above point in different directions on purpose.
• Spotify and LinkedIn statistics are self-reported by the platforms and describe their systems, not independent audits of creator income or feed fairness.