A.A.Murakami: Using Claude as a Creative Studio Collaborator

1. Who A.A.Murakami are

A.A.Murakami is the collaborative art duo formed by Anna Dumitriu and Alex May. Their practice is unusual even within contemporary art: they create large-scale immersive installations that center on ephemeral natural phenomena — fog, soap bubbles, plasma arcs, living microorganisms — and place these phenomena in dialogue with digital technology, robotics, and artificial intelligence. The result is work that is simultaneously ancient and futuristic, scientific and poetic, precise and uncontrollable.

Anna Dumitriu has a long practice working with biological materials — bacterial cultures, fermentation, the material traces of disease — and exploring how science shapes our understanding of the body and of life itself. Alex May’s background is in software, electronic arts, and interactive systems. Together, their work sits at an intersection that most institutions struggle to categorize: it is not quite science art, not quite media art, not quite installation art, but something that requires all three vocabularies to describe.

What defines their aesthetic is a fascination with the boundary between the controlled and the uncontrolled. Fog dissipates unpredictably. Soap bubbles burst at moments no formula can perfectly predict. Plasma arcs follow paths shaped by air currents and electromagnetic fields. These materials cannot be fully commanded; they can only be coaxed. This makes the act of creating with them an exercise in designing conditions rather than designing outcomes — a very different relationship to making than conventional digital fabrication, where the output perfectly matches the input.

This approach to materials has a philosophical dimension that runs through all of A.A.Murakami’s work: a conviction that interesting things happen at the edge of human control, where intention meets contingency. The work invites audiences to experience this edge rather than simply observe a finished artifact.

2. “The Moon Underwater” — the work and the collaboration

“The Moon Underwater” takes its title from a 1946 George Orwell essay in which he describes the particular pleasure of a pub that does not quite exist — a romanticized, idealized English pub with good beer, a coal fire, and no children. The moon reflected in water is the classic image of something beautiful that cannot be touched: you reach for it and your hand passes through. A.A.Murakami used this image as the conceptual seed for an installation in which the viewer stands before a body of water and sees the moon reflected in it — but the moon is inside, beneath the surface, unreachable in a new way.

The installation involves careful engineering of light sources, water surfaces, fog layers, and viewing geometries to produce the illusion convincingly. But the engineering is in service of an emotional and philosophical experience: the sensation of beauty that exists just out of reach, of nature made uncanny by its containment in a gallery space.

Claude entered the process not as a tool for designing the engineering — though it can do that — but as a studio collaborator for the conceptual development. The artists describe using Claude in a way that is closer to a conversation with a thoughtful colleague than to a query sent to a database. They explored the history of moon symbolism across cultures. They investigated the physics of reflection and why water creates such a compelling mirror. They researched precedents — other artists who have worked with water, light, and illusion — and Claude’s breadth allowed it to surface unexpected connections between works the artists might not have encountered before.

3. How artists use AI differently from developers

When a software developer uses Claude, the interaction typically follows a recognizable pattern: the developer has a specific task with a measurable output — write this function, debug this error, refactor this module — and evaluates Claude’s response against that specification. The conversation is often short. The output is concrete and testable. Success is clear.

Creative practitioners, especially those working in contemporary art, tend to use AI in a fundamentally different way. They often do not have a specification. They have a feeling, a question, a vague intuition about the territory they want to explore. They are not looking for the right answer — they are not sure yet what the question is. In this context, the value of Claude is not its ability to execute instructions but its capacity for generative dialogue: asking questions that reframe the problem, surfacing knowledge from unexpected domains, making connections the artist had not considered.

This distinction is not just a matter of how AI is used; it reflects different conceptions of what creative work is. For a developer, most of the creative work happens in the design phase before coding begins, and then the implementation is a matter of technical execution. For an installation artist, the creative work never cleanly separates into “design” and “execution” — the materials talk back, discoveries happen during making, and the final work often diverges significantly from the initial concept in ways that improve it. In this context, a collaborator who can engage with in-process thinking, respond to half-formed ideas, and offer new directions is far more valuable than one who can only execute well-specified tasks.

Claude’s particular strengths align well with what artists like A.A.Murakami need. It has broad knowledge across art history, materials science, philosophy, and physics — the combination of domains that their practice requires. It can engage with poetic and metaphorical language, not just technical specifications. It can hold a creative premise and build on it rather than immediately reducing it to something actionable. And it can be genuinely surprised — it can follow a line of thinking into unexpected territory rather than forcing every conversation back to a predefined answer structure.

4. Claude as a thinking partner — what this looks like in practice

The concept of a “thinking partner” is worth dwelling on, because it is distinct from most common framings of AI assistants. An assistant executes. A tool does what you tell it. A thinking partner does something different: it thinks alongside you, which means it sometimes pushes back, asks clarifying questions, or introduces a consideration you had not thought of.

When A.A.Murakami describe their use of Claude in developing “The Moon Underwater,” the quality they emphasize is not efficiency or accuracy — it is the texture of the dialogue. Claude could engage with the symbolic history of the moon across different cultures without reducing it to a list of bullet points. It could hold the emotional register of an idea (the longing quality of a reflection that cannot be touched) and find relevant precedents in art history that shared that quality, rather than just returning works that were formally similar. It could move between the poetic (what does it mean for light to be imprisoned in water?) and the technical (what angle of incidence produces the clearest reflection on a water surface?) within the same conversation.

This kind of dialogue — moving fluidly between the conceptual and the material, the historical and the contemporary, the emotional and the technical — is genuinely useful for artists working with complex installation practices. It is not something you can get from a search engine, which returns pages ranked by relevance to a keyword. It is also not something you get from most specialist collaborators, who are expert in one domain and less familiar with the others. Claude’s particular combination of breadth and ability to engage with context makes it well-suited to this cross-domain creative function.

The practical implication is that the most valuable use of Claude in creative work is not to ask it for answers but to think out loud with it. Share a half-formed idea and see what questions it asks. Describe a material you are drawn to and ask it what that material has meant to people in other contexts. Tell it what you are trying to make the audience feel and ask it what in art history has successfully created that feeling before. The responses will not be the final work — they will be raw material for the real creative labor that only the artist can do.

🧠 Check your understanding

5 questions · your answers are saved in this browser only

1. 1. What kind of materials are most central to A.A.Murakami's installations?

   Marble, bronze, and traditional sculptural materials Ephemeral natural phenomena like fog, bubbles, and plasma Oil paint and traditional canvas 3D-printed polymer structures

   A.A.Murakami (Anna Dumitriu and Alex May) are known for creating immersive installations centered on ephemeral natural phenomena — fog, soap bubbles, plasma arcs, living microorganisms — placed in dialogue with digital technology.

2. 2. In A.A.Murakami's practice, what is the significance of using materials that cannot be fully controlled?

   It reduces production costs It creates work that explores the boundary between intention and contingency It makes the work easier to ship and install internationally It ensures each viewer sees exactly the same thing

   Materials like fog and plasma cannot be fully commanded — they can only be coaxed. This means A.A.Murakami design conditions rather than outcomes, and the work explores what happens at the boundary between human intention and natural contingency.

3. 3. How does an artist's typical use of Claude differ from a developer's typical use?

   Artists use Claude through an API; developers use it through the interface Artists use Claude for open-ended generative dialogue and exploration; developers use it for specific tasks with measurable outputs Artists use Claude to write code; developers use it to write prose Artists and developers use Claude in essentially the same way

   Developers typically bring Claude specific tasks with measurable outputs. Artists often use Claude for open-ended exploration — when they do not yet know what the question is — valuing its ability to make connections, ask generative questions, and engage with half-formed ideas.

4. 4. What does it mean to use Claude as a 'thinking partner' rather than an 'assistant'?

   You let Claude make all the creative decisions You use Claude only for research, never for generating text You think alongside Claude — it asks questions, pushes back, and surfaces unexpected connections rather than just executing instructions You share your finished work with Claude for feedback after it is complete

   A thinking partner thinks alongside you — asking questions that reframe your problem, introducing considerations you had not thought of, and engaging with your in-process thinking. This is different from an assistant, which executes specified tasks.

5. 5. According to the lesson, which type of question tends to produce the most generative responses from Claude in a creative context?

   Closed, specification-like questions ('What are the exact dimensions of...') Open, exploratory questions ('What is it about reflected light that creates such a powerful sense of longing?') Questions with right or wrong answers Questions that reference specific dates and historical facts

   Open, exploratory questions invite Claude to surface connections, engage with emotional and symbolic registers, and introduce unexpected material — all of which are more generative for creative work than closed questions that have a single correct answer.

Reset quiz

🛠️ Build it yourself

Follow these exact steps to reproduce it yourself · estimated time: ~20 minutes

Prerequisites

• Claude or another capable AI assistant
• A vague creative concept you have been thinking about
• Curiosity and willingness to follow unexpected directions

Step 1 — Start with a genuinely vague concept

Do not arrive with a fully formed idea. Start with something unresolved — a material you are drawn to, an emotional quality you want to produce, an image that keeps coming to mind. Write it to Claude in a few sentences without explaining what you want to do with it yet.

Example: “I keep thinking about the moment just before a rainstorm, when the light changes and the air smells different. I am not sure what I want to do with this but I keep returning to it.”

Step 2 — Resist asking for ‘the answer’

When Claude responds, notice if you are tempted to immediately ask “so what should I make?” Resist that. Instead, pick one thread from Claude’s response that resonates and pull on it. Ask a follow-up question about that specific thread.

Example: “You mentioned that many cultures have rituals for the moment before rain. Can you tell me more about a specific one that you find particularly striking?”

Step 3 — Ask for three unexpected connections

At some point in the conversation, ask Claude: “Make three unexpected connections between this idea and things I would not have thought to look for.” The connections should feel surprising. If they do not, push Claude: “Those feel too obvious — can you go further?”

Step 4 — Research precedents in art history

Ask Claude: “What works in art history — installations, performances, paintings, anything — have successfully created the emotional quality I am describing? Not formally similar work, but work that produces a similar feeling in viewers.” This surfaces reference points that can inform your own work without copying it.

Step 5 — Reflect on what the dialogue produced

After the conversation, read back through it. What did you discover that you did not know before you started? What question did you not know you had? What is the most unexpected thing Claude said? The value of the dialogue is often visible in retrospect more clearly than during the conversation itself.