Why Does AI Make Only Some Scribblers More Creative?

AI doesn’t make you more creative. It exposes how you already think. If you use AI to avoid thinking, to skip synthesis and judgment, you’re training yourself to be a Matrix battery.

Dan Porter, CEO of Overtime, posted some vibe-coding insight on LinkedIN: “for the tech-adjacent crowd, the people who understand how software works but can’t actually code, that’s where the unlock is. You can get close to building something meaningful.”

I commented about vibe-coding retro games and basic tools, how it taught me what devs face when they’re hunting bugs and living through spec changes. But scribbling is my domain. I understand how words work like a dev understands code. Maybe that explains why my AI experiments are useful to me more often than not. And freak me out re LLM’s ever-increasing power.

Many use generative AI. Brainstorming, outlining, summarizing, rewriting, compressing, expanding. Creative results are all over the map. Some scribblers feel sharper, faster, more playful. Others get the same weak sauce answers on repeat, just delivered at light speed.

I don’t think the gap is about talent and lame AI. It’s about how you approach using LLMs.

My LLM Awakening

After my first couple hours with ChatGPT, I had an epiphany: much of my career involved operating like a large language model for the showrunner.

They’d come into the room with a general direction. We snacking scribes would pitch options and generate drafts. The boss would rework everything into alignment with their voice and intent. We were the pattern-matching engine. They were the filter.

That existential revelation got me over my trepidation. AI wasn’t just an inevitable threat. It might be a tool that could amplify my process the way a room of staff scribblers amplified showrunner vision.

Over the past few years, I’ve been experimenting with ChatGPT, Claude, Midjourney, Gemini, ElevenLabs, Perplexity, Etc. They’ve become part of my workflow, not because they execute a final product I think is rad, but because they bring a fresh angle I can bounce off.

The trick for me has been treating LLMs like a room full of scribblers, not an oracle.

Your Brain Matters More Than Your Prompts

Organizations rolled out AI expecting a creativity boost by default. More ideas, better ideas, quicker breakthroughs. The reported creative payoff has been mixed. One widely circulated summary of recent findings cites a Gallup result: only 26% of employees using generative AI report improved creativity.

That tracks with what many of us feel day to day. The tools work. The results can be uneven. So what separates the folks who are getting creative support vs. those who don’t?

Not better prompts. Better judgment loops.

The scribblers who benefit from AI don’t treat output as a final draft. They treat it as raw material. They think about their thinking while they iterate. Planning an approach, monitoring whether it’s effective, noticing when they’re legit stuck or just accepting something because it’s there, then adapting tactics instead of blindly hammering ahead.

Research published in the Journal of Applied Psychology (field experiment, 250 employees) backs this up: AI improved creativity mainly for employees who used metacognitive strategies. They actively tracked and revised their thinking, using the tool to gain “cognitive job resources.”

Scribblers sans metacognition hit a wall. They ask, AI answers, they move on. No beat to ask:

- Is this right for my story?

- What assumption is baked into this response?

- What version of this idea would freak me out a little?

W/O reflection, AI’s a shortcut generator. The result could be competent and clean, but it’s probably lifeless. This is why some scribblers feel faster but trend flatter after heavy AI use.

Scribble Room Model

When I scribbled for Alias, I served as the room’s action and espionage specialist. I drew from my fandom of the genre, video game literacy, and deep diving the subject in college. That’s the right mental model for working with LLMs.

- Break big problems into smaller, focused questions.

- Frame questions with context only you can provide. - Iterate with follow-ups that sharpen constraints. - Filter ruthlessly with professional judgment.

If you understand a domain deeply, you can ask better questions, spot generic moves, recognize when a suggestion violates the “rules” of the arena, then push the tool into more interesting territory.

Creativity comes via human filters, not machine mind generation.

My ADHD-Friendly Workflow

ADHD keeps my brain spinning like the flux capacitor, churning out more ideas than my prefrontal lobe can hold. LLMs help me generate versions quickly so I can see what’s sticky and what’s toss-able. This helps me fail faster, so I can decide what to explore.

Because I hate staring at a blank screen, I built a workaround. Years before LLMs arrived, I started compiling my craft knowledge into question sets. Every time I hit a story problem or solved one in the writers’ room, I’d capture it as a diagnostic question.

- What does this character want right now?

- What belief are we testing?

- What’s the emotional turn in this scene?

These questions became my Scribbler’s Checklist, reminders of what I knew about crafting my scenes, sequels, characters, and stories.

When LLMs showed up, I already had my frameworks locked down. I fed that material into an LLM and had it generate development questionnaires. The checklist gave me a foundation. The LLM gave me speed. Together, they let me prototype story architecture without the blank-page paralysis that used to kill my momentum.

The work I’d already done organizing my thinking made effective LLM collaboration an adjacent possible. Sans that scaffolding, I’m just another dude yelling questions at a robot.

Dictation Over Perfection

Don’t open a chat and beg for ideas. You’ll get something, but it probably won’t be what you need. A classic newb move is giving AI a blank page and expecting it to read your mind.

Bring your obsessions. Pet theories. Half-baked takes about why zombie movies play better in winter. Your prompt is a map for someone who’s never been to your neighborhood. The more landmarks you share, the less likely they end up at the wrong house.

Prompt A: “Give me heist movie ideas.”

Prompt B: “I think every heist movie is secretly about found family. The job is the excuse to throw strangers together until they care about each other more than the money. Give me three plots where the heist fails but the family succeeds. Think Ocean’s Eleven meets The Remains of the Day.”

I often start with dictation, blathering into Docs, Voice Recorder, ChatGPT, whatever gets me off page one. I’ve found AI can take my garbled ADHD notes, organize them, suggest clarity fixes. This helps my head-canon get readable to folks not living inside my skull.

Architect vs. Bricklayer

Every AI response is like a first draft from an eager intern who’s watched a thousand movies and never left their bedroom. Tons of raw material, limited lived experience and emo insight.

Let AI handle:

- Fifty title options.

- Rough outlines from your premise plus beat brainstorm.

- Untangling clunky sentences.

- Variations on tropey phrasing you can punch up later.

Keep for yourself:

- Your core argument.

- Character choices that matter.

- The emotionally specific moments that land with humans.

If I’m not rescribing at least two thirds of what comes back, I’m absolutely avoiding the work. You can be the architect and let the tool lay bricks faster than you can scribble solo.

My Research Intern

LLMs are great for tracking down specific details that can take hours, sometimes days. AI compresses that search time. The catch is accuracy. These tools can confidently hallucinate. Most of my work is fantastical hoo-ha, but if you’re scribbling about anything real, you’d better get some verification. Treat every factual claim as total BS until multi-source confirmed.

Learning Via Failure

A couple years back I tried using Claude to turn an unsold screenplay into prose to publish as a novella. It wasn’t a one-click wonder. It was a full-on grind. The result kind of sucked. But when I ran a few chapters through ElevenLabs text-to-speech, the audiobook version sounded cool.

That failed experiment taught me about the diff between scribbling for the screen vs. the ear. The output wasn’t the win. My accelerated learning curve was.

The Risk Going Forward

The danger isn’t that AI will replace scribblers. The danger is that scribblers stop practicing the skills that keep them worthy of the trade, and their taste tuned-up.

Early research and reporting illuminates how heavy LLM reliance reduces cognitive engagement, and may weaken recall. This is a warning worth respecting.

Scribbling is mental and emotional weightlifting. Remove resistance and the muscles go soft. Keep your creative abs strong. If you’re not rewriting a meaningful chunk of what comes back, you’re letting the LLM do your reps.

AI doesn’t make you more creative. It exposes how you already think. If you use AI to avoid thinking, to skip synthesis and judgment, you’re training yourself to be a Matrix battery.

It’s critical to protect all the hard earned knowledge gleaned from your unique experiences. If you’re a scribbler, you understand storytelling the way devs understand coding. Use that wisdom to frame sharper questions and critique outputs. Lather. Rinse. Repeat.

Don’t let these powerful, and terrifying tools vibe you into forgetting your hard earned skillset, unique taste, and emotional insight.