The creative workbench#

How Karlie uses AI and digital tools inside a creative practice

Purpose#

This document should show the creative process from the artist's perspective.

The important question is not whether AI “made” something. It is how Karlie uses dialogue, reference, generation, selection, rejection, composition, and iteration to develop judgment and make work. Taste appears in the decisions: what is pursued, refused, combined, changed, or recognized as finished.

This material may also help MOOI NIET. understand how an artist's process can contribute to a collection's presentation and a host's experience without becoming an exhaustive explanation or content obligation.

Karlie's account of the workbench#

What belongs in the workbench#

Which tools, conversations, materials, references, physical actions, and forms of attention participate?

What AI contributes#

What does it make easier, stranger, faster, more expansive, or newly possible?

What remains Karlie's judgment#

Where do selection, refusal, taste, authorship, correction, and finishing live?

What should remain private#

Which parts of the process need protection in order for experimentation to remain free?

What might enrich a host's experience#

Which fragments could deepen the work without telling the host what to see?

Creative-session entry#

Copy this section for each useful conversation or making episode.

Session: short title#

Known facts

• Date:
• Project or work:
• Tools involved:
• Source notes:

Starting impulse

What was Karlie trying to find, make, test, or understand?

Workbench process

How did dialogue, generation, reference, physical making, or iteration unfold?

Taste decisions

Record concrete acceptances, refusals, corrections, combinations, and changes of direction.

Result

What physical, digital, conceptual, or experiential artifact emerged?

Presentation possibility

Could any part of this process deepen a host's encounter? What should remain absent or secret?

Karlie's words

Preserve selected excerpts or a first-person account reviewed by Karlie.

Agent synthesis

Optional and explicitly labelled.

Questions that may emerge across sessions#

• How is taste encoded through repeated decisions rather than stated preferences?
• When does dialogue create possibility, and when does it flatten the work?
• Can selected process fragments become part of a living work or collection score?
• How can MOOI NIET. expose a creative lens without producing a Wikipedia page about the artwork?
• What does a host gain from seeing a path not taken?

Candidate account from the painting-review process#

Agent synthesis — draft for Karlie's review

The painting practice uses AI and digital tools as part of a larger chain of attention: generate or gather possibilities, select among nearby options, paint from a chosen source, reflect immediately and after time, then update the criteria for what to seek next. The important authorship lives in the decisions—what is pursued, refused, simplified, translated, or allowed to remain unresolved.

The review system deliberately separates the experience of making from the quality of the final result. It asks what stayed interesting, what became dead, what was difficult in a useful way, whether the image held its mood, and whether the source environment mattered. Patterns are updated only after several entries, so a single success or failure does not become doctrine too quickly.

The practice also contains a refusal of premature explanation. In the active expression mode, feedback is limited to one or two reversible experiments, while interpretation, symbolism, quality judgments, and audience-based correction are held back unless invited.

Concrete sessions#

Session: Painting review system#

Known facts

• Date: 2026-04-12
• Tools involved: ChatGPT, image-generation references, physical painting, Markdown archive
• Source notes: 2026-04-12--painting-review-and-expression-method.md

Starting impulse

To identify which source-image qualities lead to paintings that feel enjoyable, successful, and worth continuing.

Workbench process

The system records a source before painting, the painting experience, the immediate result, and a later reflection. It compares what was selected with nearby options and turns repeated evidence into recurring patterns, reject patterns, and future prompt adjustments.

Taste decisions

The documented criteria favor mood-worlds over literal subjects, environments that can simplify into strong value blocks, emotional restraint, ambiguity, and subordinate secondary elements. Clutter, over-explicit information, and descriptive overworking are recurring rejection signals.

Result

A reusable review structure and an archive of provisional criteria.

Presentation possibility

Agent synthesis: A host might be offered a trace of selection or a path not taken, but the full studio process should not become an explanation obligation.

Unknowns

The source note summarizes local project documents; the original conversational exchanges that produced every rule are not all available in this contribution.

Session: Expression, lightly guided#

Known facts

• Status: active working mode
• Source notes: 2026-04-12--painting-review-and-expression-method.md

Starting impulse

To keep expression flowing without turning each painting into a message, series, or performance for an audience.

Taste decisions

The working mode protects gesture, momentum, mass, and stopping signals. It favors reversible physical experiments and limits feedback to one or two actionable suggestions.

What remains Karlie's judgment

Whether a mark strengthens or subdivides a mass, whether a painting still reads at a distance, when to stop, and whether something feels alive or dead.

Unknowns

The available document describes the operating stance, but not a single dated session with a fully documented result.