AI and Our Collapsing Creative Horizons
Everyone's talking AI. Some of it is hype. Some of it is fear. The real risk for creatives sits underneath both: a steady compression of originality.
The danger isn't robots replacing you. It's your taste, your process, and your voice getting trained to think inside a machine's averages. That's how creativity gets flattened - fast, quietly, and with your consent.
The Great Flattening is already here
Film, music, and publishing were drifting into sameness long before AI. Smaller risks, bigger franchises, playlists engineered for skip-proof background noise. Now add tools that reward speed, imitation, and "good enough."
Lower-level scripts get "polished" by prompts. Music charts fill with mimicry built from reference tracks, not lived experience. Reading is replaced by summaries that strip out tension, subtext, and the struggle that makes a work stick.
Pattern remix isn't originality
Ask a model for a "private eye searching for a missing woman," and you'll get plot beats stitched from a thousand detective tropes. It can be clever. It can't be truly new.
These systems output the statistical center of what already exists. Depend on them, and your work drifts toward the median. Convenience smooths edges. Edges are where your style lives.
How flattening shows up in the work
- Ideas feel familiar before they're finished. You've seen that twist, cadence, or color palette a hundred times.
- Process turns into prompt roulette. Less sketching, less rewriting, fewer "bad takes" that grow into breakthroughs.
- Voice gets outsourced. The model's rhythm starts replacing your own.
- Audiences adapt to summaries and scraps. Full works get treated like optional context.
The uncomfortable truth
AI gives you what you think you want, quickly. That's the trap. By skipping the grind, you skip the parts that refine judgment, deepen taste, and produce work with a pulse.
As Sam Adler-Bell put it: "The things that enrich my life - art, books, movies, human relationships - are, thus far, being made worse, cheaper, uglier, by this technology." The convenience tax gets paid in soul.
A practical code for creatives
- Draft before you prompt: Write the ugly first pass yourself. Sketch before references. Record scratch vocals before you comp anything.
- Keep AI out of concept and style: No "write in the style of X." Study masters directly. Steal like an artist, not like an autocomplete.
- Use it for grunt work, not taste: Transcripts, cleanup, alt titles, proofreading, spreadsheets. Taste stays human.
- Collect raw inputs from life: Field notes, interviews, location photos, messy voice memos. Feed your model - your mind - with reality.
- Build a voice canon: Document 10 principles of your style (syntax, pacing, motifs, values). Use it as a checklist, not a prompt.
- Impose craft constraints: Word counts, limited palettes, live takes, one-take scenes. Constraints create texture machines can't fake.
- Read whole works, slowly: No summaries for core influences. Depth compounds. Skims don't.
Safe use cases (that won't sand your edge)
- Organization: Summarize meetings you already had. Tag research you already gathered.
- Exploration: Generate questions you hadn't considered, then do the thinking yourself.
- Technical lift: Code snippets, formatting, file conversions, storyboard placeholders - temporary, not final.
- Quality control: Consistency checks against your own style canon, not the model's defaults.
Guardrails for originality
- Ratio rule: At least 80% of creative labor stays human. Reserve 20% for assistive tasks.
- Provenance rule: If you can't explain where a choice came from (a lived moment, a study session, a draft), it doesn't make the cut.
- Repetition test: If the output feels instantly familiar, push it further or throw it out.
Signals to watch in the industry
- Budgets shift from talent to tooling; teams lose apprenticeships and the pipeline of future masters shrinks.
- Executives optimize for "consistent performance" over singular voices; IP farms beat originals.
- Audiences expect more, pay less, and binge faster; depth loses to volume.
For context, not comfort
Adoption data shows why this is moving fast, even if the outputs feel same-y. See the Stanford AI Index for scale and trendlines here.
If you choose to skill up with AI, do it with intent
Learn prompts as constraints, not crutches. Study workflows that keep taste and decision-making human while offloading admin.
Curated resources by job can help you set those boundaries: Complete AI Training - Courses by Job. For research-focused learning and critical perspectives on AI, see AI Research Courses.
The line worth defending
Tools can speed you up. They can't give you a soul. Protect the parts of the process that hurt a little - that's where voice is forged.
Keep the machine in the back office. Keep the art at the front of your life.
Your membership also unlocks:
