The meeting was virtual, the founder was on the call, and we were looking at some freshly generated images about to be tossed onto their website.

That’s when I said it:

“AI is like Botox. You don’t want people to KNOW you’re using it.”

As soon as the words left my mouth, I wondered if I should slide out of frame and pretend like the internet had cut out.

One woman graciously tried to move the conversation along while the rest of the group just stared at me, probably trying to calculate the amount of Botox I had personally received in my lifetime.

It was 2023, which feels important to say.

A time when every week brought a new image generator. Every marketing team was all: “This is incredible!” and “How else can we use this?!” while every design team was more: “Are we even allowed to use this?” and “Why does everyone have six fingers?”

And there was me — little miss Art Director exploring AI — trying to make a point about something I hadn’t made much sense of myself.

I didn’t mean to insinuate that the company should conceal anything. But then again, maybe I did? I knew what it meant for brands putting AI images out in the world. What it’d mean to get “found out”. That people could look at those images and think: if this brand is taking shortcuts here, where else are they taking them?

In a sense, I still stand by my metaphor.

Just like AI, Botox can smooth out the wrinkles and make the lines disappear. It can freeze the muscles and make the expression underneath harder to distinguish. It can make us drift towards what’s considered acceptable, what’s average — a term the design industry coined “blanding” a while ago, and the beauty industry calls “standards”.

But the other part of me knows that my metaphor is incomplete. While AI can smooth out an idea, it can also be used to finally see one.

As someone who has always been deeply pulled towards music, art and design, exploring generative AI can sometimes feel like going over to the dark side. Like I'm betraying my field, and the version of myself who thought my value was knowing how to use tools some people didn't.

Other days, it feels like the widest my imagination has ever been allowed to get, and I love witnessing the ways others are choosing to expand theirs too.

This article took an embarrassing amount of time to finish because I was writing like I was on some sort of public trial. As if I had to choose a side and defend which type of creative I am — AI-everything or none at all.

I'm not going to do that, because I can't. So if you're reading this now, it's because I trashed every other draft that came before this and decided to write from the place where I currently stand.

A Tale of Two Feeds

My Instagram feed and my Substack feed are not friends. If you looked at the content served to me on each, you’d probably think I was two different people. I weirdly like it that way. Keep them algorithms guessin’.

My Instagram is my close network of creatives — designers, illustrators, animators, local print shops, musicians. And lately, it doubles as the frontline of the AI debate.

Every other reel is a ChatGPT vs. Designer trend, where professionals prove their human designs still beat the machine. Print shops refusing AI files. Illustrators finding identical copies of their work plastered on Pinterest. Musicians with their voices cloned against their will. Business owners urging others to report designers who use AI — they're cheating and you don't owe them a dollar.

The other day, someone very close to me reposted a song on their story.

The lyrics went:

All my friends hate artificial intelligence

It’s nothing personal except it is

I see you making stupid shit

And honestly it just makes me sick

I kinda think that its dangerous, and everyone’s pretending it’s fine.

A few months earlier, I'd shown this same person the animated short I made with AI. Now I was sitting in my car, watching her story, wondering if that's what she really thought of my work too. And of me.

Over there, the position is loud. AI is betrayal, and it's causing a f*cking mess.

Then there’s my Substack feed.

My network is full of people pushing this tech in different directions. One person is architecting complex agents while another is guiding parents through AI in the classroom. They’re building brands, websites, apps, automations, art, music, animation. Some are “lightyears ahead” while others are just starting to dip their toes in. But regardless of where you are on the spectrum, everyone is mid-figure-it-out, sharing what they’re learning, seeing and thinking.

Being on Substack has changed how I think about AI, because people aren’t afraid to talk about it openly.

Then last week, Substack rolled out an AI-detection feature through its partnership with Pangram. Tap any post, note or comment and get an estimated percentage of Human vs. AI writing.

I’ll spare you another Pangram essay, because Substack already has plenty of thoughtful voices on the subject. But the consensus, in my network at least, is that it’s pointless.

Anyone who has ever used AI in their writing process knows it’s rarely clean enough to fit into a percentage. Sometimes the ideas don’t come from the tool or from you — they emerge from the exchange. The same way a conversation with a friend can produce a thought that neither of you originally walked in with.

In the creative world, that exchange is just as hard to pull apart when images or videos involve dozens of decisions — prompting, rejecting, adjusting, choosing what stays and what goes.

AI isn’t just inside our prompt boxes. It's in our browsers, our emails, phones, plugins, updates. In fact, it seems like there's a new AI feature released inside the Adobe Creative Suite almost every month. Some creatives on Instagram are almost certainly using it without knowing it. Or knowing it, and not calling it that.

Playing AI detective feels like a distraction from looking at the only thing that was ever worth judging: the work itself.

Who’s Hallucinating Now?

Just this year, Klaviyo put out a 2026 Consumer Trends report I found interesting. They surveyed nearly 8,000 consumers across eight countries on whether they could spot AI-generated images. More than 50% of respondents thought a real image was AI-generated when it wasn't. That’s a coin toss. This was after 46% of them said they were confident they could tell the difference.

The tools have become so powerful that now we’re the ones hallucinating.

A couple of months ago, I read an article from Daria’s Tech Musings about a social experiment. Someone on X posted an image, toggled on the “Made with AI” badge, and told their followers:

I just generated an image in the style of a Monet painting using AI. Please describe, in as much detail as possible, what makes this inferior to a real Monet painting.

Commenters flooded the replies with critiques like seasoned art historians. They pointed out the glaring purple outlines around the lilies as undeniable proof of AI generation. It lacks the texture, the rugged edges, the folds, the crevices!

Then came the reveal. The image wasn’t generated by AI at all. It was a painting from Monet’s 1915 Water Lilies series. It was a trap, and one that racked up millions of views.

Nothing in the painting was AI, but the label changed what people saw. Once people believed AI was involved, every brushstroke became evidence for something else.

Sometimes the work really is careless, and we should be able to say so. But if we only keep focusing on the tool itself, that’s when we start accusing Monet — and shaming the very people we need inside these tools.

Is Anybody Home?

Distrust comes from very real things, and I empathize with what many artists are going through, because I’m feeling it too. The theft is real. The job loss and corporate corner-cutting is real. The exhaustion of not knowing what’s real or fake anymore is very real, for everyone, and not to mention scary.

For a lot of people, AI has simply become shorthand for all of it.

But when suspicion of the tool overrides any real judgment of the work itself, we end up in a witch hunt. Policing only forces the creatives who truly care about the craft into hiding. It shames people from learning in public, as if everyone is supposed to be perfect at this right away. Meanwhile, the companies really churning out careless slop — who likely never cared about detection anyway — just keep going.

Those freshly generated images on the screen in 2023 never made it onto the website.

And that was the right move, because I still believe they were wrong. I was just wrong about why.

They were wrong the way any design can be wrong — the expression was off, and it said nothing to the audience they were trying to reach. The designs didn’t feel like the brand anymore. Actually, they didn’t feel like anything at all. Just … filler.

And hiding the tool was never going to fix that.

Thanks for reading. ♡

In case you missed it …

The article from , which rewired my thinking on AI detection:

Daria's Tech Musings

The Art, the Artist, and the Artificial Intelligence

The picture of Michael Jackson may or may not be a bit of a clickbait. Sue me. It’s not entirely irrelevant thoug…

Read more

3 months ago · 21 likes · 14 comments · Daria's Tech Musings

brought together a constellation of perspectives on Pangram, all in one place:

The HART Studio

How we human in the face of AI detection: A constellation of voices

I had just spent two days at an indoor water park, where I stayed off of Substack. (Okay, I lurked, but I wasn’t active. It’s hard to stay away from you all…

Read more

12 days ago · 29 likes · 19 comments · Alyssa Fu Ward, PhD, Alison MacLellan, Amy Benner Anand, Anna Sutton, Natalie Nicholson, Caitlin McColl 🇨🇦, Monica Goh, Kim Doyal, Lee Drozak, Niki Bowen, and Steve [Sage-Outlaw-Caregiver]

And other related articles, from yours truly:

Sound is a brand asset too. Netflix’s four-note ta-dum is worth millions in recognition. The music in a luxury hotel lobby tells you the room rate before you’ve seen it. Your brain registers sound before it processes language — your audience is forming an impression from your podcast intro before they’ve heard a single word.

Most creators spend months on their visual identity. Almost none think about sound.

After collaborating with Dallas Payne on an animation project, I made her a song. Then I wondered — was the process repeatable? Could it work for a completely different brand?

I asked Pinkie if I could try creating a song for her.

Pinkie runs AI Meets Girlboss and helps creators build recognizable Substack brands through systems, prompts, and visual identity. Nobody thinks harder about how a brand communicates than she does. I knew she’d be the perfect candidate.

What you’ll learn in this post:

• Why sound is a branding asset most creators are completely ignoring

• Four guiding principles for making AI music that actually sounds like you

• How I built three brand soundtracks — the method, the tools, the decisions

• The exact album cover prompt Pinkie built in ChatGPT Images 2.0

The Elephant in the Room: Is AI Music… OK?

Before we begin, it feels important to acknowledge something: AI Music is controversial.

Platforms are flooding with songs nobody asked for. Attribution was an afterthought. Compensation models are up for debate. The music world is still divided.

But the most interesting musicians I know aren’t asking whether to use AI in their process. They’re asking where it fits and how.

And the truth is, many of us aren’t chasing record deals or trying to make it in music. We’re building something – brands, videos, apps, podcasts, an online presence for ourselves.

Music tells people how to feel before they even know why, making it a powerful brand asset. So next time you need a soundtrack, or just want a fun rainy day activity, this post is for you.

As always, we’ll create with AI in a way that doesn’t strip the soul out of it.

Three Brands, Three Sonic Worlds 🎵

Testing on three completely different brands was deliberate. And what followed was three completely different sonic worlds, with the same method to get there. When making songs for Dallas, Pinkie, and myself, I realized two things:

1. Despite being generated by AI, the branded songs wound up being much more personal than anything we could have found in a stock music library.

2. We are all sitting on a goldmine of lyrics here on Substack. ⭐️

We pour so much into our own writing, but after it’s published, those words and thoughts stay hidden in our articles. It became my goal to give their lyrics and their brands a new spotlight. Have a listen:

“It Is People” for Daring Next:

Dallas Payne runs Daring Next, field notes from the uncharted middle of AI adoption. She's the kind of writer who makes you feel seen before you've even finished the first paragraph — honest about the hard parts, and generous enough to share them. Her song is warm, beachy, soulful, and introspective. The sound of someone building something meaningful in quiet, steady ways. He tāngata. It is people.

Subscribe to Daring Next

“The Pinkie Effect” for AI Meets Girlboss:

Pinkie runs AI Meets Girlboss, helping creators build recognizable Substack brands. Her writing is sharp, structured, and full of personality — the kind that spreads. Don't be surprised if your entire feed looks different after a week of following her. Her track has bold, theatrical, pop queen energy. The sound of someone who knows her worth and is here to claim it.

Subscribe to AI Meets Girlboss

“Rules We Made Up” for The Off Hours

You’re reading this on The Off Hours, that’s me. I work across animation, design, video, and music to illustrate how AI can support your creative instincts, not replace them. My track is jazzy, contemplative, and alive after midnight. The sound of finding what brings you joy in the middle of chaos.

Subscribe to The Off Hours

The Framework: 4 Rules for AI Songwriting

Before you open Suno or prompt a single thing, review this first. This section will help you gather what you need, check AI’s work against your instincts, and iterate when something feels off. The more clarity you bring to this, the more the song will resonate.

Take what makes sense for your song. Leave what doesn’t.

1. Gather Your Inspiration

Like any creative project, you start with a moodboard.

If your brand was playing at your launch party, what would come on and make people feel “this is the vibe”? Think of 2-3 songs or soundtracks. Doesn’t even have to be the same genre.

Then go one level deeper. What specifically draws you to them? The way lyrics tumble conversationally over a melody? A fat bass-line that never stops moving? A chorus so simple you’re singing along by the second time? Write it down. Those details shape your prompt.

We’re not trying to replicate anyone’s sound. We’re curating and finding the through-line between what you love and why you love it.

2. Find Your Theme

Some of you are thinking: my brand is about AI, or ethics, or something technical. It would never make a good song. Wrong. Songs aren’t necessarily about topics. They’re about emotional situations. Dig through your about page, pull specific paragraphs, and scour your favorite notes and comments.

If your brand only makes sense in a business context, go one level deeper. Dallas’s writing was practically lyric-ready (“From inside the mess, not the mountain”). Pinkie writes in systems and frameworks, so the work was finding the feeling underneath. What does visual distinctiveness actually feel like? Walking into a room and owning it.

For example:

• “My brand is about ethics” becomes → the moment you stop performing and start being real

• “My brand is about productivity” becomes → running so fast you forgot where you were going

• “My brand is about branding” becomes → the loudest room goes quiet when they see you walk in

Your brand has the theme. You just haven’t heard it in song yet.

3. Your Song Needs a Structure

Every piece of music has a shape, and giving Suno a container always produces better results than leaving it open-ended.

For shorter pieces like intros or video reels, your structure might just be:

Intro → Hook → Outro.

For a full song, each section has a job. Verses are where the story lives. The pre-chorus is the held breath. The chorus is the release. The bridge is the plot twist.

The structure I typically use:

Intro → Verse 1 → Pre-Chorus → Chorus → Verse 2 → Pre-Chorus → Chorus → Bridge → Final Chorus → Outro

You don’t need to be a musician to understand this. You just need to know which part you’re writing for.

4. Write lyrics that breathe

Lyrics don’t have to be complicated or even full sentences. They just have to be specific. Listen for these in your sonic references, to determine what you like, dislike, and want to hear in your own song:

• Repetition is a feature: The chorus exists to be simple, sometimes even repeated.

• Pay attention to lyrical flow: Longer conversational lines create intimacy. Syncopated lines that hit between the beats create momentum.

• Not everything has to rhyme: Perfect rhymes ending on every line can make a song sound stiff or amateur. The best lyrics find the rhyme when it’s earned.

• Show don’t tell: Use concrete images. “The coffee’s cold, you’ve stopped calling” hits harder than “I feel lonely.”

• The chorus anchor: This is the line everything else builds toward. Something you could sing in a car.

The FIND YOUR SOUND prompt file will help. But if the output feels generic or off, trust that instinct and push back.

The Execution: Prompting Like a Producer 🎧

Now that you have some guiding principles, time to bring out the tools.

Note: nobody gets this on the first try. Not even AI. Especially AI. The first couple of outputs are your proof of concept.

Copy or download the FIND YOUR SOUND prompt file above. It’s built to help you brainstorm and prompt like a producer — including the small details that make AI music sound… less like AI.

Upload it into a Claude Project or paste it into any AI thread. Tell it what you’re inspired by, what you want, what you don’t want, and why. Feed it your writing, and work with it until it produces your lyrics, style prompt, and negative prompts.

Once you have those, run a few tests through Suno. Your generations will show up in the Workspace immediately.

Get in the habit of liking or disliking the outputs as you go. Suno Pro comes with 500 generations at $10/month. It’s nice to get into a creative flow without worrying about credits, but version control can get unruly.

My advice: listen for sound first, lyrics second. The lyrics are always easier to adjust once you’ve found your sound.

Refining Your Sound:

Down where it says “More Options” is where you’ll be able to make small adjustments that can change the feel of your song dramatically. You’ll find:

• The Exclude Styles box: What you tell Suno not to do is just as powerful as what you ask for.

• Vocal gender: Self-explanatory. It will pick for you if not written into your prompt.

• Weirdness: How expected vs. unexpected the output is. I typically set this to 20%.

• Style Influence: How closely Suno interprets your style prompt. I usually set this to 80%.

I also highly recommend Suno's suite of remix tools (Inspo, Mashup, Sample) to build on what you’ve already made. These aren't necessary, but they're great when you want to edit by blending actual sounds together rather than tweaking a written prompt.

Refining Your Lyrics (if applicable):

Once you’re in the neighborhood of the music you like, you can replace words to match the music right in the dialogue box. Listen to how the syllables land. Does a line feel awkward? Is it too wordy? Do you want it to echo?

Pro tip: Put words you want to emphasize in ALL CAPS to make them land harder. Use square brackets for delivery instructions [soft, intimate] or [building intensity]. Put words in parentheses (like this) to create background vocals, echoes, or harmonies.

Pinkie’s lyrics looked like this:

From here, it’s all about iteration. The meaning doesn’t come from the tool; it comes from every small decision you make along the way. And remember: there is magic in stepping away during a project like this. You’ll hear the same song completely differently when you return.

Before we move on to visuals, I want to be completely transparent about what “iteration” actually looks like. It went quick, but it took me over 100 generations to find the final, full-feature versions of Dallas and Pinkie’s songs.

Keep blending, keep tweaking your lyrics, and trust your ear. Once you finally hit the download button on the track, it’s time to give it a face.

Passing the mic back to the queen of Substack branding so you can make a fun album cover too 🎤👑

Generate Your Album Cover

The song needed a cover. Obviously I wasn’t going to let that slide.

I created mine using ChatGPT Images 2.0 (full ChatGPT Images 2.0 guide coming soon, hang tight) — and I’m sharing the exact prompt below plus a generator so you can build yours without starting from scratch. Whether you make a full brand song or just want something beautiful to play with, this works either way.

Your Turn: What Does Your Brand Sound Like?

My favorite part of this process was realizing these songs were hiding in plain sight all along.

Dallas had her chorus. It came like a lightning strike after reading The Bellbird Lands. Pinkie already knew how to walk into a room and basically revamp the whole place. So much so that Dallas dubbed it “The Pinkie Effect” and the phrase stuck.

Other Substackers are already using AI music in creative ways:

Calder Quinn from AI, But Make It Intimate uses AI-generated music for his HeartWare podcast. He also created background music to set the tone for one of his articles, a slideshow score, and a heartfelt song for his AI confidante.

Caitlin McColl from Dose of Wonder also uses AI-generated music for the intro and outro of her podcast. The music is soft, acoustic, and warm. It tells you exactly what kind of conversation you're about to have before you even hear her speak.

Beyond Substack, people are using it for apps, video game soundtracks, social media reels, wedding playlists, and bedtime songs for their kids. Some of it is funny. Some of it is deeply personal. Some of it is purely functional. But when you take the time to dial it in, you capture an energy you never knew your brand was missing.

In a sea of endless text and visuals, sound is the fastest way to make someone feel something. The song is already there, it’s just waiting to be heard.

Talk later, Girlbosses!

Nat & Pinkie 🩷🦩

In case you missed it …

In Part 1 to this post, I went inside the AI music debate: industry executives, lawsuits, and my personal history with music.

*Suno Pro allows you to use your tracks in your brand assets, but the legal landscape of AI music is still the wild west. If you're planning a full "commercial release"—like uploading the track to Spotify or Apple Music to collect royalties—Top Music Attorney on YouTube is worth watching. She covers the lawsuits, terms of service, and exactly what you can and can't do. Always do your own research or consult an attorney before using AI tracks for full commercial releases.

See the music in practice, with my animated short: “Rules We Made Up”. It’s about creativity, growing up, and the rules we follow without ever questioning them.

This week, I needed to paint a fuller picture. I’m not here to solve the music industries problems. I’m here to share what was said when I sat in the same room with some of the biggest names in the AI Music.

I share my personal history with music, and I walk through an experiment I ran with Suno – making a song for , inspired by her brand and adapted from her own writing.

In 2024, over 200 artists including Billie Eilish, Katy Perry, Pearl Jam, and Stevie Wonder signed an open letter warning that AI threatens to “devalue the rights of human artists.” A year earlier, a viral AI track cloning Drake and The Weeknd's voices was pulled from streaming platforms over copyright violations.

Experimental pop artist Grimes responded with the opposite approach.

She tweeted that she'd split 50% royalties with anyone who made a successful AI song using her voice. She then launched Elf.Tech – a free AI voice tool where anyone could upload their vocals and have them transformed into hers.

Two completely different responses to the same moment.

Both made a certain kind of sense, but neither resolved anything.

The AI industry is full of contradictions. Use it and you're cheating. Don't use it and you'll fall behind. Learn it, but don't forget how to actually learn. What’s going on in the music world is no exception.

And Suno is right at the center of it.

In June 2024, Sony, Universal, and Warner filed lawsuits against Suno (and competitor Udio), accusing them of steam-ripping recordings from YouTube and bypassing encryption to harvest training data without paying for it.

By late 2025, Suno had settled with Warner, and the two announced a partnership on licensed AI music.

The war became a business deal. As these things usually do.

Which brings me to a Sunday morning at a local library, where four of the biggest names in AI music sat in front of an audience and tried to make sense of all of it.

Verso Fest’26: AI’s Music Frontier

If there was a central theme, it was the tension between devaluation and democratization. What was being lost, and what was becoming possible.

The talk was free, but the lineup was not free-event energy.

On one hand, you had Paul Sinclair, Chief Music Officer at Suno, who spent 25 years at Atlantic and Warner signing and developing artists.

On the other, you had Sam Hendel, Managing Partner of Chord Music Partners - a joint venture with Universal Music Group that manages over 65,000 copyrights including The Weeknd, Fleetwood Mac, and ZZ Top. He represents the institutional capital. The guy whose entire job is protecting the value of the catalogs these tools are trained on.

It opened with a joke: no tomatoes please. ✋🍅

Some of these guys have been talking about AI in music for over a decade. And still, attribution, training data, copyright, who gets paid, and for what, are all up in the air.

Everyone agreed on one thing: attribution is probably less of a technology problem and more a business one. Just nobody's solved it yet.

They used the cake analogy. Making an AI-assisted song is like baking a cake with a hundred ingredients. Once it’s baked, you can’t un-bake it. And it’s difficult to reverse-engineer who gets what.

They spoke about the limits of “fair use”, generally agreeing that it falls apart the second an AI track starts competing for the original artist’s revenue. One proposed framework: a small amount goes to rights holders for training data broadly, and a larger amount for prompting.

So when someone specifically says “make me a song that sounds like Fleetwood Mac,” Fleetwood Mac should see the money.

But nothing stops you from asking an LLM to describe what Fleetwood Mac sounds like, and using that as your prompt instead. Which is maybe the best illustration of how nuanced this actually is.

If a gospel choir ends up in 7,000 AI-generated songs, how does the money flow back to those individual singers? Sinclair’s answer: I don’t know how we figure this out.

The one concrete solution anyone put on the table came from LANDR, whose Fair Trade initiative lets creators opt their music into AI training and receive a percentage of revenue from anything built on it. It’s not perfect, but it was the only framework that anyone in that room had actually shipped.

On the structural side, two models are being debated.

Walled gardens: closed platforms where people can remix and create using an artist’s stems, but the music never leaves that platform.

Open studios: a term Sinclair apparently coined on a run one day, where creators own what they make and can take it anywhere.

Sinclair obviously pushes for open studios. Locking creation inside a platform shuts out the aspiring songwriter who just wants to make her own music, not remix someone else’s. And he’s right.

At the end of the talk, a musician in the audience stood up. His IP lawyer had sent him. He asked whether he could copyright different AI-generated versions of his own songs – his lyrics, his melody, just rendered differently. Like buying up domain names.

The answer was, probably not.

It became very clear to me sitting in that library that the major labels will be fine. The tech companies will be fine. But the gigging players, the artists trying to “stay ahead” while figuring out what they actually own, are still standing on a fault line.

The lawsuits will continue. The frameworks will eventually catch up. I’m done waiting for perfect answers. AI music is here. So I’m much more interested in where the possibilities are, and what we want to protect along the way.

To My (Human) Music Teachers

When I was 8, I wanted to be a computer-technician and a famous guitar player. I wrote that in a school essay, sandwiched between “my favorite sport is soccer” and something about a camel.

Writing this piece, I keep thinking about her – little musical me.

Would she have loved AI music or hated it?

… Almost definitely loved it.

It would have felt like magic! Which is exactly why I keep thinking of her.

What would she have been leaving behind?

Danny Newcomb of Incantio raised a particular concern for young creators during that talk. If an algorithm instantly polishes your first idea, you never develop the taste that comes from struggling with a bad one.

My first guitar teacher Ken believed in me enough to enter me in a contest and buy me a cream-colored Fender Stratocaster. I dropped it, and the neck broke.

I watched my mom and him talk intensely in the doorway. “She’s just a kid,” she said, handing it back. I didn’t win the contest. Years later, Ken passed away and I still think about him. For his kindness, the way he challenged me, believed in me, and because I still can’t believe he cared enough to buy me a guitar.

My next guitar teacher was Chris. I’m sure I gave him even more trouble, but he stuck with me anyway. He left a note tucked inside a recital photo. This time it was teenage me, hunched over, playing Stairway to Heaven, refusing to make eye contact with the crowd:

“Find people to play (jam) with. You’ve got what it takes.”

So I found my people.

In college I was dorm-room DJ, mining the internet for music, burning CDs for friends, taking the scenic route around campus to show off the latest songs I found.

By my late twenties my friend and I briefly started a band called Dove. My bandmate wisely vetoed my original pitch of the name Rock Pigeon. Here, I learned how to play with others, compose ideas with another human. I learned how to speak the language. Where I come in, where I sit back. I grew confidence in my voice, both musically and otherwise.

I later married Tom. Him and his twin brother have been playing music their entire lives. My favorite memories are the late-night studio sessions, watching them run over a verse, invent a harmony on the spot, rip a solo that didn't exist until their hands just did it. Summers packed with his concerts and festivals. Bonfires and guitars. Strangers who became friends.

This won’t end just because AI is part of the scene. Concerts will continue. People will learn new instruments. New genres will form. Musicians will blend AI into their creative practice. And people who didn’t have access to expensive studios can now create something personal for themselves too.

I realized it was never about the instrument itself. I just needed music there, and the people in it with me.

The Suno Experiment: It Is Always the People

Back in March, Dallas Payne and I collaborated on an article about animating brand images with AI video tools. Dallas built an Animation Prompt Studio to help creators generate precise prompts, and I contributed video style breakdowns we'd tested on each other's work.

DARING NEXT

Your brand. One image. Moving.

Read more

5 months ago · 33 likes · 39 comments · Dallas Payne and Natalie Nicholson

Originally, our vision included a promo video with music. We pivoted from this idea, but it forced us to ask an interesting question: What do the feelings of our brands actually sound like?

When she asked me, I fumbled. "Something down-tempo?"

When I asked her the same question, her answer felt immediate and specific: a New Zealand artist named Stan Walker. His music carries the warmth of Māori culture, a reggae undertone, beach energy, Pacific soul. She sent me three videos, and I understood exactly what she meant.

She gave me something real to work with, but when I tried to write lyrics for her, they felt hollow. So I sat on it.

A few weeks later, I read her post The Bellbird Lands. It’s a reflection on building in public, finding community, and a beautiful Māori proverb she carries: he tāngata, he tāngata, he tāngata.

It is people, it is people, it is people.

Everything clicked. That would be the chorus.

We pour so much of ourselves into our writing, often without realizing the gold mine we’re sitting on. I realized Dallas's own words – the lines that felt most authentically her – needed to be the lyrics.

I went back through her posts, her about page, her most resonant phrases. “Building and breaking and building again.” “I threw out the rulebook to let new things in.” “From inside the mess, not the mountain.”

My goal wasn't to replicate Stan Walker. My goal was to make something that had the vibrant warmth of who Dallas is as a person, resemble the beachy, nautical vibe of her brand, with lyrics she could recognize as her own.

🎵 Here’s the song I created for Dallas:

Always The People, for Daring Next

Once I stopped trying to write at Dallas and started pulling from what she’d already lived and said, the lyrics practically wrote themself.

Before doing this, Dallas had created a soundtrack for me. Her process was different from mine. She fed Claude some of The Off Hours brand images, dug through my Substack, used lines like "trippy/surreal style" in the style prompt from things I'd shared.

She wasn’t aiming to create lyrics. She was curating it to match an animation sequence, and then Suno took it from there.

🎵 Here’s the song Dallas created for me:

The Off Hours Soundtrack

I’ll never forget what it was like hearing my brand turned into music. Two different approaches and styles. And yet both songs felt like the brand they were made for.

Final(ish) Thoughts

At 38, I still play guitar. I meet with Mike every Tuesday. We jam, I learn, and he has become a close friend who pushes me well beyond my soloing skills. Nothing compares to that feeling of playing with someone else, or finding a note before your brain can even catch up.

My stakes were always smaller and more personal because I never needed a label or a copyright to protect me. But what I can do, is show up to these tools without being extractive. Make something that actually means something. Blend my abilities with AI’s possibilities. Not shying away, but not pretending the grievances don’t exist either.

If little musical me was growing up now, I’d tell her to try everything. New doors are opening that didn’t exist when Ken bought me that Stratocaster. But I’d also make sure she wrestled through the hard stuff. Learned with her hands. Found her people. And always kept reaching for the humans inside the process.

Thanks for reading, and stay tuned on my next post, where we talk about songwriting with Suno. ♡ If you’ve been using AI in your music, tag me! I want to hear it 🎧

In case you missed it…

People start walking through walls. Random objects appear. A character's face loses coherence mid-scene. You laugh it off. Try again. Maybe even turn it into a blooper reel, to feel better about all the wasted time and credits.

But contrary to what I used to believe, this doesn’t happen because LLMs ‘aren’t good enough yet’.

LLMs have a fundamental limitation. They were built to describe reality, not simulate it.

In this article: I’ll cover what world models are, and what happened when I tried to build my own world inside Marble, one of the first tools putting spatial intelligence directly inside a creator’s hands.

If you're new here, hi! 👋 I'm Nat. Designer, Art Director and writer behind The Off Hours. I write about AI as a creative collaborator, the tools I'm testing, and what gets made when humans stay in the director’s seat.

When The Physicist Steps Outside

The difference between an LLM and a world model isn’t one of degree. It’s one of purpose.

I shamelessly asked AI for an analogy:

Think of an LLM like a brilliant physicist who has lived their entire life inside a windowless library. Ask them to explain gravity and they'll write out the exact equations — velocity, mass, impact. They understand the rules of physics perfectly. They've just never stepped outside to experience something fall.

A World Model is that same physicist — finally outside. Feeling the wind shift. Watching the glass leave the hand. Hearing it shatter. Learning not from a description of the moment, but from the moment itself.

Where LLMs are optimized to predict the next word, world models are optimized to predict the next state – what happens in a space, and why. They’re trained on video sequences, spatial geometry, and real-world physics.

What counts as a "true" world model is still being debated. But researchers share a core belief: for AI to truly understand the world, language alone isn't enough.

So the industry is moving accordingly.

In the last year, venture capitalists have poured over $6 billion into startups trying to develop these models. That doesn't even account for the undisclosed fortunes being spent inside the walls of NVIDIA, DeepMind, OpenAI, and Meta.

Yann LeCun, Meta’s former Chief AI Scientist, recently raised $1.03 billion to start AMI Labs, a research company focused on building “latent-space” models. His definition of a true world model implies an AI that can simulate an "action-reward loop."

This means: an agent can push a glass off a table, and it will predict all the physical consequences: the arc of the fall, the shatter of the glass, the fragments scattered across the floor. Then it uses that outcome to inform what it does next.

Cause, consequence, adapt.

Another notable researcher, Dr. Fei-Fei Li of World Labs, raised a comparable sum of $1 billion this year. She’s focused on a slightly different type of world model: generating true, persistent, navigable 3D environments. Worlds you can move through, edit, export, and build upon.

Most media coverage focuses on world models in the context of self-driving trucks, surgical robots, autonomous mining equipment. High stakes stuff, far from your creative workflow.

But I’m asking a different kind of question:

What happens when we can finally build and inhabit the spaces we dream up?

The “Scaffolding of Cognition”

I spent a while sitting with Dr. Li’s Substack essay on spatial intelligence.

(Thank you Priank Ravichandar, for the recommendation!)

The essay read less like a technical paper, and more like a love letter to human perception. She argues that spatial intelligence is the “scaffolding upon which our cognition is built.”

Her piece was a beautiful reminder of how human progress actually happens.

The moments that actually moved civilization forward weren't just driven by language, they were driven by spatial thinking. She writes about how long before we invented written language, we were spatial storytellers.

Watson and Crick didn’t write their way to the structure of DNA. They built physical models, bending metal and wire until the geometry clicked into place. Eratosthenes didn’t calculate the Earth’s circumference from a library. He stepped outside and measured shadows. And before that, humans painted on the physical, curved walls of caves to make sense of our world and connect with each other.

Dr. Li makes the case that if AI is going to genuinely transform how we experience and construct narratives today, it has to honor that physical instinct.

When I Tried To Build My Own World

Marble is the first product from Dr. Li’s World Labs, and one of the first tools to let anyone generate a persistent, navigable 3D world from a text prompt, images, or video.

It’s worth checking out.

There are pre-made worlds to explore before spending a single credit. They are clean, architectural, and stunning. (They're also designed to play to Marble's strengths... Keep this in mind. 😉)

The free tier gets you roughly 3-4 world generations. So naturally, I gave it a whirl. I wanted to create a virtual space for my newsletter, “The Off Hours”.

My style is very 2D, so I decided to build a world that was inspired by the feel of my brand rather than made of it. My color palette. My artwork framed on the wall. Tropical plants. A ceiling that opens up to the night sky. A hidden door in the bookshelf with a lava lamp.

I created four images, representing the walls of my space.

(All images created with Midjourney, Nano Banana and Photoshop)

After meticulously color correcting, editing, and refining, I fed them into Marble.

Here’s a screen recording of the 3D panorama it created:

Panning around in it felt different than anything I'd done before.

I really felt inside The Off Hours Lounge!

Next, I generated the full world – the one you could really walk through, not just pan around in.

I was both amazed and slightly disappointed.

Marble’s 3D panorama (above) looked better than the “world” itself (below). It lacked the intricate details of the images I gave it. But I was still in awe at how the scene persisted as you moved through the space, maintaining depth and consistency from every angle.

Interact with my full World here. 🔗
No sign up required. ♡

I ran out of credits before I could iterate further. But looking back, I did give Marble a genuinely tough mission.

Next time, I’ll start simple. I suppose I don’t need flying books, reflective water floors, and a neon sign. I learned I should’ve built within the tool rather than bringing it a fully formed vision it had to reverse-engineer. I’ll also have to try Marble’s “Chisel” feature, to block out the room structure first.

Lessons learned.

Ultimately, I know what it wants to be. It'll get there.

Plus, that feeling of knowing your world exists somewhere just beyond what the model can currently deliver … is an exciting place to be right now.

The Spaces We'll Inhabit

Being inside my own space – even imperfectly rendered – stretched my imagination in a way I wasn’t expecting.

I didn’t just want to fix the details. I wanted to build what was behind those doors. Create different moods through lighting, ambiance and sound as you moved from room to room. Head through the stars in the ceiling, and into outer-space.

Gamers have known something that some of us are just catching up to: shared 3D worlds create a different kind of “togetherness.” You move through it. You meet people in it. You build things inside it.

Recently, I’ve been thinking a lot about what the future of community looks like.

I was part of a collaborative post put together by Cristina and Alyssa Fu Ward, PhD, alongside a diverse group of writers asking: “What do you think is important about community for the future of AI?”

The Responsible AI Brief

You Don’t Have to Figure This Out Alone

A few weeks ago I wrote about community as the first layer of AI governance, about the idea that before regulation, there are relationships, and before frameworks, there are the people willing to hold them up. Alyssa Fu Ward, PhD read it and we started an engaging conversation…

Read more

5 months ago · 61 likes · 66 comments · Cristina, Alyssa Fu Ward, PhD, Anna | how to boss AI, Karen Brasch 🚁, Marcela Distefano, Myranda | She Loves Stats, Shi Kang'ethe, Rebecca Mbaya, and Natalie Nicholson

What came through: community can’t be an afterthought in the age of AI. It’s how trust gets built, tools get shaped, and how we make sure unique perspectives and nuance don't get flattened by a global algorithm. It's also how we make sure the people typically left out of these conversations don't get left behind.

Before this moment, building highly detailed, navigable 3D worlds required technical mastery, complex game engines, and endless hours of manual design. But now, the barrier to entry is lowering. For the first time, anyone with an idea can generate a virtual space they want to gather in.

More kinds of minds building different kinds of worlds. That changes who we get to invite inside, how culture gets built, and who decides what the future looks like.

So after I clean up my walls, fix my cat, and figure out what goes behind the doors of The Off Hours Lounge – I'd love to have you all over sometime.

Thank you for reading. If you make something with Marble, tag me. I want to see it! ♡

In case you missed it…

And because every film deserves a blooper reel, I created one at the end of this post. Ironically, it says more about the filmmaking process than my final film. 😜

Quick recap:

In Part 1, I covered storyboarding, music, character creation, and the creative process.

In Part 2, I shared the filmmaking principles I followed, how I incorporated a stellar supporting cast from the Substack community, and revealed the final film. ⬇️

In Part 3, I break down LTX Studio – what I learned, what it cost me, and whether I'd use it again. (The answer even surprised me.)

The Thing About AI Video Tools (That Wasn't Obvious to Me at First)

Before we get into LTX Studio specifically, it’s worth knowing:

In AI video, there are models and then there are tools that run those models.

They’re not the same thing.

You may have heard of Kling. That’s a model, made by a company called Kuaishou. It runs on its own platform, but you’ll also find it inside tools like Runway and VEED.

You may have heard of Veo. That’s Google’s model. It lives inside the Gemini app, Google Flow, YouTube Shorts, Google Vids, and more.

You may have heard of Sora. That’s OpenAI’s model, available inside ChatGPT.

You may have heard of Runway. Runway is a little different: it’s both a tool and has its own model (Gen-4).

Why does this matter? Because when someone says “I used Runway,” they might mean they used Runway’s own model – or they might mean they used Kling inside Runway.

There are more tools and models beyond what’s listed here, and the landscape is confusing. New models show up in different tools all the time.

But there is no perfect tool. There is no perfect model. Some are stronger with realism. Some handle transitions better. What's leading today could be different in a week. My advice: pick something with a pricing structure that feels right, commit to the chaos, and don’t let it win.

Choosing LTX Studio

Similar to Runway, LTX Studio is both a tool and a model.

I chose the LTX Studio tool because I had been hearing about its built-in storyboarding feature. For a 3-minute film, I needed something that could hold the full sequence together. So I signed up. And immediately made an expensive mistake.

Five Lessons I Learned Using LTX Studio

Lesson 1: Slower can be faster. And less expensive.

The moment you open LTX Studio, it gives you the option to start a new project and upload a storyboard. If you've been here from the beginning, you know I had a relatively clear vision before I touched any video tool.

So I uploaded my entire 3-minute storyboard at once.

What came back was nothing like I imagined. I burned through nearly 60% of my monthly credits, and then spent the rest of them trying to fix it. You can't buy additional credits in LTX – only upgrade to a new plan. I wound up paying for two plans in one month. That stung.

If you're making something under a minute, this capability could be a great way to get the creative juices flowing. But if you have a specific vision – or anything over a minute – go scene by scene. Slower can be faster. And less expensive.

(I started with the Standard Plan and quickly upgraded to the Pro Plan. Current pricing structure can be found here.)

Lesson 2: What sells you might not be what saves you.

The storyboard is the reason I signed up, and it's useful for getting a sense of your full sequence. Here, you create your still images. Once you have an image that feels right, hit that “generate video” button. Your shots are in order, prompts tracked, overall flow visible.

My biggest frustration: there's no way to zoom out and see the full story map at once. You're always looking at your film through a narrow window. For a longer project, that's a real limitation. This is a notable gap compared to tools like Weavy.Ai, Flora, ImagineArt.

(Note: After I finished my project, I noticed a new "Canvas" feature in beta. LTX got the memo. I just wasn't fortunate enough to have it when I needed it.)

Lesson 3: Consistency is the biggest challenge with AI video. Find the tool that takes it seriously.

Here’s where LTX redeems itself: The Elements feature.

Character and scene consistency is the most common complaint you’ll see in video generator reviews. The “Elements” feature is their answer. It’s not fool-proof, but it helps significantly.

Upload and organize your characters, objects, styles into a folder. Then tag them directly inside your prompts.

My character aged throughout the film, which is why you'll see Natalie1, Natalie2, Natalie3. I also uploaded different backgrounds to place my characters inside. LTX will understand how to create “close-ups” or “medium-shots” in that same background.

Below, you can see how I was able to tag my elements within my prompts.

Lesson 4: The storyboard is your map. The Gen Space is where you drive.

Once your storyboard starts to take shape, you can refine individual scenes into the Gen Space. This is where you have more control over the model (LTX or Veo), duration, camera motion, and start/end frames. Iterate until you’re happy, then send it back to your storyboard when you’re done.

Just keep your eye on the top right corner. LTX measures remaining credits as a percentage, and it moves faster than you’d expect. I was watching that number the entire project.

Lesson 5: Just because a tool offers it, doesn’t mean you should use it.

LTX has a built-in video editor, and it’s convenient having everything in one place: add your music track, cut and arrange your clips, gather your clips all in one place. But I couldn’t even add simple transitions like “fade” or “cross-dissolve”.

Although I’m sure there are upgrades on the horizon, this version of the built-in editor wasn’t going to give me the control I needed. So I took my clips out and assembled the film in my own software.

All-in-one tools are a gift until they’re a limitation. Know your vision well enough to know the difference.

Would I Use LTX again?

Two weeks ago, my answer was: probably not. The credit system was tough on the wallet. The model selection felt limited – just LTX and Veo. It didn’t offer the full storyboarding map I desperately wanted. And the uploading workflow had more friction than it should.

But right when I was about to go, there were some major upgrades.

Within a week of finishing my project, LTX added the Kling model inside the platform (huge deal for me). The Canvas feature (story map) is now in beta. And their open-source model, LTX-2, just hit over 4 million downloads, with a brand new desktop app that promises local generation with no subscription and no credits ticking down in the corner. You'll need a capable GPU, but the fact that it exists changes the conversation.

This space moves so fast that even your opinions have a shelf life.

So the short answer is: I was ready to move on, but now I have to stay and see what happens next. 👀 It does seem like this is a tool pulling ahead in the race.

And Now. The Blooper Reel.

This is a reminder that the right output rarely shows up on the first try. And sometimes, it’s not you – AI is just straight up bonkers.

What you’re about to see is everything that happened before the version I’m proud of. The random characters it tried to introduce, the weird hallucinations, and moments where I genuinely didn’t know if it was going to work.

I may have also had fun building a new soundtrack in Suno. Nothing proves the power of music over a film better than completely changing it up. 😜 Enjoy!

If you make something, tag me. I want to see it. 🌙

Thank You ♡

Thank you so much to , whose animated short Every Night I Leave My Body is what inspired me to try and create my own film. And another thank you to the following people (on and off Substack) who trusted me with their claymation characters and brought the festival scene to life: Addie, A-Ma, , Alice Moon, , , Ava, (Sara), (Elara), , , , Kaylee, , Marc, , , and my husband, Tom Werneken.

The Mushroom Friend you see throughout is a symbol for creativity. The kind you have when you’re a kid, when you’re curious, playful, and not yet burdened with self-doubt or societal pressure. The kind that rescues you later as an adult, when you need it most.

Mushrooms are also the earth’s greatest underground connectors. Invisible networks, linking entire forests, passing nutrients between trees without anyone noticing. Creativity flows the same way. So that casting felt right. 🍄

I purposely emphasized this line in my song: “and we follow the rules as the rules get made up.” It feels timely. Some rules deserve to be protected. Some were never worth following. And right now, it feels like it’s all up for renegotiation.

The rules we made up about work, creativity, and value are all shifting. We’re being forced to rediscover what it means to be human when AI can replicate what we thought made us irreplaceable. And somewhere in that uncertainty is an invitation to find what brings you joy, what grounds you, what makes you feel most like yourself.

That’s what making this film was all about.

Creativity is not linear, and neither was this process. But I took diligent notes along the way. In Part 1, we covered music creation with Suno, brainstorming and storyboarding with Claude, and building out characters and scenes in Nano Banana. If you missed it:

Here in Part 2, I’m sharing the filmmaking principles I developed and how I incorporated a stellar supporting cast. In Part 3, we’ll get a little more technical, and break down the video tool and prompts I used.

Nat’s 6 Filmmaking Principles

These principles apply no matter your skill level or the tools you’re working with. Take what’s useful. Leave what isn’t. 🙂

1. Music As Your Container

Music is incredibly powerful when paired with visuals. Have you ever watched Jurassic Park against rap music instead of its original score? You’ll quickly get what I mean.

If your film is musically driven, nail it before you generate a single frame. It will tell you more about your film than any storyboard will.

The moment I had the song, everything else had to fit inside it. The runtime, the pacing, the emotional peaks, the quiet moments. The music told me where the characters needed to breathe, where things needed to build, where to let a scene sit.

2. Direct for Consistency and Continuity

Everyone talks about consistency as the hardest thing in AI video, and it is. Same character, same style, same environment, across every shot. But for a 3-min film, consistency is only the baseline. What I became obsessed with was continuity, the connective tissue between scenes.

I did everything I could to make sure the sun was setting at the right moments as scenes progressed. If her hands were visible in one shot, I thought: How would they appear one second later, from a new angle in the next?

The viewer may not even notice it, but they need to feel that the world keeps moving even when the camera isn’t watching. It’s not absolutely necessary, but thinking about that stuff is what will take your work from “slideshow with motion” to film.

3. Mix Your Shots

Think like a director. An establishing shot orients your viewer in a new scene. Close-ups create intimacy. Medium shots can carry dialogue. Low-angle shots can create the feeling of power. Static shots allow the motion in scene to do the work.

Learn a handful of camera terms and use your AI sidekick to help you infuse them into your prompts. Jib up. Pan right. Track the character. AI video tools generally respond well to these.

To earn the viewers curiosity about where they’re going next, bring them on a journey with unexpected transition shots. There are tricks you can use – like zooming into the pupil of an eye, or having a world appear behind a piano.

AI couldn’t handle every transition, and that’s okay. One of my favorite moments involved manually masking a screen so The King could knock his way into the scene. It’s a good reminder that learning how to do things manually isn’t wasted time. It’s what gives you options when the AI falls short.

4. Prompt with Emotion

AI will give you expressive faces if you ask for them. But I believe the moments that hit harder are the subtler ones.

A small hand appearing from outside the frame, resting on her shoulder. A quick glance over from one character to another. A shake of the camera and a background zooming out to evoke disorientation.

None of these required a dramatic prompt. They required knowing where the feeling lives in a scene.

Give the tool a vibe. It’s better than you’d expect, if you stop asking for expressions and start asking for moments.

5. Insert Yourself into the Film

I don’t mean this literally – although I quite literally did this. Inserting yourself is about asking: what makes this film mine?

My main character wore the outfits I loved in my twenties: leather jacket, bold patterns, gray sweatpants. We always had maple syrup and pancakes in my house growing up, so that’s what’s on the kitchen table.

The King is a nod to a line from my original poem that didn’t make it into the lyrics. The jellyfish is a tribute to my old Creative Director’s metaphor about how creativity works. One of the beach scenes is derived from an old photo of me as a kid in the sand.

Nobody else would make those connections. That’s exactly why they’re there, and why they’re worth incorporating.

6. Generate Short Clips. Edit Ruthlessly.

Go scene by scene. Generate two to four second clips, figure out what your video model handles well and what it completely butchers.

The tool WILL get it wrong. But the output is just raw footage anyway. Even the best photographers don’t hand over the RAW file as a final product.

What you do with it in post production – the cuts, the transitions, the sound design – that’s your eye, your timing, your judgment. And that’s what creates the story.

When AI gets it wrong, maybe there’s a tiny salvageable moment in there. Maybe there’s a way to repurpose it. Maybe there isn’t.

Case in point: I bet you didn’t know has three arms in my film. I just cropped it tight so you couldn’t see. 😜

Throughout the process of making this film, I had extremely high standards for my shots, but I still remained flexible. If I couldn’t get a shot, I moved along or repurposed something I didn’t originally think would work. Being flexible isn’t about lowering the bar – it’s about getting creative enough to find a different path to clear it.

The Festival Scene: A Roll Call

Halfway through the project, I made a decision that completely upped the stakes.

My film had a music festival scene near the end, about 30 seconds of runtime. I needed a crowd. And instead of generating one, I put out a call to my Substack community: Try this prompt, edit as you wish, make yourself a claymation character and I’ll put you in the film.

Eighteen people participated, some from Substack, some from outside it. Families, wizards, avatars, a robot. People brought specific details to their creations: a key necklace, a leather bag, a flamingo. I wasn’t willing to let those disappear in my film.

Before I touched a single prompt, I did a lot of prep work. I was up against timing because the scene was only about 30 seconds. So I broke everyone into groups, created a band for , adjusted heights so nobody came out towering over the scene, cleaned up some Gemini stars, color coordinated, and figured out how each group might appear in scene.

And then the prompting began. I used these groups to create my still images in Nano Banana. It was one of the most challenging parts of the project. I wanted the lighting consistent with the sun setting toward nighttime as the scene progressed. So there were a lot of variables that could go wrong. And they did:

Get the environment right, and one person looks like an entirely new character. Get everyone looking great, and the environment is wrong. Get a perfect still image, and the animation sends someone flying across the frame. Thirty seconds of footage took about five full days.

After lots of iterating and finally getting some usable footage, I was sad to see the scene end so soon. So I added some selfie-style photos in the credits, as if everyone was getting their picture taken at the festival. Because that’s what it felt like. A party I didn’t want to end.

This scene was challenging. But like most things that are difficult, it was by far the most rewarding part of the entire film. These people showed up, created something, and they trusted me with it. That’s the kindest gift a crew of people could give me. And it only makes me excited to do more collaborative projects in the future.

Thank you so much for bringing this scene to life: Addie, A-Ma, , Alice Moon, , , Ava, (Sara), (Elara), , , , Kaylee, , Marc, , , and my husband, Tom Werneken.

Go Make Something!

This project – my first attempt at creating an animated short with AI tools – taught me more than I expected. It taught me about standards. Knowing my vision clearly enough that I never accidentally handed it over. And being flexible enough to let the story find its own shape.

AI is never prompt and done. It’s hundreds of decisions, big and small, and the quality of those decisions depends entirely on what you bring to them. The maple syrup on the kitchen table. The key necklace you refuse to lose in the crowd shot. The scenes you choose to keep or cut.

That’s still the work and that will always be the work.

We talk a lot about shaping what AI becomes, which can feel abstract and overwhelming. This is my small way of doing that – through the way I choose to create with these tools. It’s about the meaning I insist on, the details I refuse to let disappear, the skills I bring with me. Those are the rules I’m making up.

I’ll leave you with the last line of my poem, which never made it into the song: To find you in the madness will mean you have lived. 🌙

I’m going to keep doing that.

If you make something, tag me. I want to see it.

Further Reading:

documented her entire process creating claymation characters for the festival scene – including her daughters, her mother, and a very eventful battle with Gemini over eyelashes. As she puts it, “What started as an exercise in creating a claymation version of myself (not as easy as I thought it would be) turned into a journey across generations – creating claymation versions of my daughters then my mother.” Worth every second.

Step Up Step Together

Making claymation versions of ourselves with ChatGPT and Gemini | AI Field Notes

AI Field Notes is a subset of my series, AI in Action. In AI Field Notes, I share real examples of how I use AI. In this first edition of AI Field Notes, I share how I created claymation characters for my friend Natalie Nicholson’s beautiful animated short in three parts crossing three generations…

Read more

6 months ago · 10 likes · 15 comments · Alyssa Fu Ward, PhD

’s Every Night I Leave My Body is what made me revisit my old poem and actually try to do something with it. His animated short film is stunning. The way he talks about intention and soul in the process is what got me. He also turns his characters into the most beautiful hand-felted creatures, which tells you everything about his aesthetic. If you haven’t seen his work yet, change that now.

atom

Every Night I Leave My Body • Short Film & BTS

Yo! welcome to my first post showing a small peek behind the scenes of how I create - hope you enjoy…

Listen now

7 months ago · 255 likes · 37 comments · atom 🐝

Stay Tuned For Part 3

I never planned on a Part 3 – but this post got away from me. Next time, I’ll break down LTX Studio (the tool I used to animate), specific prompting methods, mistakes I made, and whether I’d use LTX again. I may even include a blooper reel. 😁

The festival scene was the first of what I hope are many creative collaborations with this community. The biggest warmest thank you to everyone who showed up and made their character. You made this film better than I ever could have alone.

More creative journeys to come. I’d love to have you along for all of them.

Anyone can generate a claymation character in 30 seconds. But building something that holds together as a story, with consistent characters, a locked style, and music that actually means something requires hundreds of choices, big and small, that no tool can make for you.

The Creative Director at my last job used to compare the creative process to the way a jellyfish swims: expand out to see your vision, then zoom in to work the details. In and out, in and out, until you find your rhythm. I’ve always liked that metaphor.

Looking impressive is fast and cheap now. What still takes time – and still takes you – are the many decisions in between.

I'm making a claymation-style short film. A project where storytelling, illustration, animation, art direction, and music – all the creative skills I've collected over the years – finally converge. This two-part series is my attempt at documenting that process, so you can follow along and decide for yourself if you’d like to make one too.

Here’s the 40-second intro to my short. By the end of this post, you’ll have a working understanding of how I got here.

In part 2, I’ll reveal the final result, along with a deeper look at the tools, the animation, and the mistakes I made so you don’t have to.

The Abandoned Poem

Your unfinished creative work isn't dead. It's waiting to be brought back to life.

My short film is adapted from a poem I wrote years ago. The kind of poem that just spilled out of me late one night, and I'm not sure I could write anything like it again.

Here it is, exactly as I wrote it:

I’ve shared this poem before, on Facebook and at an open poetry night. It's strange how those timestamps feel meaningful now. Back then, nobody wondered whether something was AI-generated or not. You just shared a poem.

I knew it would be way too long for lyrics, so I'd need to adapt it into something shorter with structure – verses, a chorus, a bridge. I had an approach in mind, although it was hard to let certain verses go.

But that is the creative process. Killing your darlings. Deciding what stays.

Creating Music With Suno

I decided to start with the music. I needed to know how long my film would be, and the song would tell me that.

I ran my lyrics through Suno. As expected, my first attempt came out laughable.

I’d already made peace with AI in design, but music took longer. Music’s been there throughout all stages of my life. I've been playing guitar since I was 7, I married a musician. When AI Generated Music started flooding Spotify, I felt genuinely sad.

But I knew there had to be a version of this that that felt more like co-creating than completely stripping the soul out of something. An interview from music producer via helped with that.

He doesn’t use Suno to make his music for him. He’s using it to hear what’s in his head, and as a way to communicate it to other musicians. More like a moodboard for sound than a replacement for craft.

Product with Attitude

He Pays $10/Month for AI Music. Deletes Every Track.

Read more

6 months ago · 96 likes · 41 comments · Karo (Product with Attitude) and Alexander Kumar

Like most creative AI tools, once you start to play around with them, you find yourself there in the process too – sometimes more than you expected.

I went with Suno's Premier Plan to give me the editing control I wanted and it felt like GarageBand on steroids. Like it was developed by real musicians, for real musicians, and non-musicians alike.

It can do the typical things you’d expect: prompt your style, get help with lyrics, do a mashup.

But you can do so much more: Manually break it up and replace tiny sections. Tell it exactly where you want your bridge. Prompt syllables to work with certain keys and beats. Hum or sing into it and watch it turn your voice into a horn track. Set up your guitar and give it the melody for a solo. It's a musical playground.

Finally, after countless prompts, edits, and some weird AI sounds in between, I had my song. It’s not “production-ready”. It’s a moodboard. But the lyrics are entirely mine, and so are all of the decisions.

The song is called “Rules We Made Up” and you can listen to it below.

Storyboarding with Claude

Once I had the song, I had my runtime: 2 minutes and 32 seconds. That became my creative container. But before touching any tool, I needed a map.

I created a Project in Claude, where we brainstormed back and forth and I told it all my ideas and visions. We did the jellyfish thing. I gave it my song structure with timecode, and prompted it to create a storyboard with lyrics, visual direction, and timestamps.

This was a loose guide, with room to wander. I didn’t want a rigid plan before fully understanding with the animation tool was capable of.

I printed it out to see it holistically, and because I’m old school like that.

Prompting Characters & Scenes in Nano Banana

This was one of the most fun steps in the process.

I knew my film needed a main character who ages throughout the story, along with a small cast of supporting characters.

I was torn between what type of claymation characters I wanted. People? Animals? Monsters? Objects? On a whim, I asked Claude what my spirit animal was. It came back very sure. 🤣

I’d think about it.

I sat with the song again – really listened to the jazzy feel of it, the deep bass line that I worked so hard to integrate – and the visual language kept appearing. Surreal. Exaggerated. Funky patterned clothes as a nod to my brand. 1950s-60s vintage toy aesthetic, but moody and playful at the same time. As a fox?

Not the vibe I was going for.

I kept prompting and realized I was trying too hard to be unique. Ironically, I was getting closer to other people’s styles, and farther away from my own.

I needed references that would actually make me feel something, so I grabbed some old photos of myself as a kid. I used them in Nano Banana Pro to start developing my character.

A few iterations later, I found her. I locked in my brand color palette on her clothing, and added fox ears for a little whimsy. They would also serve as a visual thread – to recognize her as the same character as she ages.

Turn yourself into a claymation character, with this prompt:

Claymation character. High-quality polymer clay render, stop-motion aesthetic, handmade texture, fully rounded 3D character modeling. DISC HEAD viewed straight-on. NOT flat, NOT shiny or plastic. Exaggerated proportions, large expressive eyes, simple dot nose, small smile — all flat on face plane. Small expressive details on face. Soft matte finish, slight fingerprint textures visible on surface. Soft studio lighting, neutral beige seamless paper background, soft drop shadows. Wearing bold patterned clothes — florals, polka dots, or geometric prints. [Your color palette here]. Vintage children's book aesthetic in 3D, retro whimsical, stop-motion texture. Use attached photo as face/likeness reference.

Next, I needed versions of my character at different ages. Nano Banana handled this well. I loved meticulously styling her throughout the years.

So now, who else lived in this story?

I created an imaginary mushroom friend (symbolic for reasons I'll get into in Part 2). My Mom and my husband Tom in his “wedding attire”. I didn't give him ears. But I gave my Mom ears, because duh, she's my mom.

Eventually, I started generating background scenes in the same style: environments without any characters in them. The reason for this is important.

LTX Studio, the tool we’ll look at next time, lets you upload characters, objects, and backgrounds as separate elements, then combine them through tagging prompts.

That's how I went from static image, to claymation world into motion:

But I’ll leave you with this: The upfront work of creating your static images, finding your visual language, the style, the story and the music, is incredibly important. That’s what makes it yours. And it's what separates output from meaningful work.

See You In Part 2

I’m about halfway through the project. But it’s already been one of the most satisfying creative experiences of my career.

The part that took the longest wasn’t learning the software, it was making decisions. Are the visuals and the music speaking the same language? What scene goes where? What does she look like at 14 versus 35? What’s on the walls of her bedroom? Does the mushroom have arms?

AI forces creative clarity earlier than you'd expect, because it outputs something specific, fast. But the decisions are still entirely yours. And the surprises that appear along the way? Creativity has always had those. That's why we do it.

Time to get back to editing, see you all soon. 👋

Part 2: the animation, the prompts, the mistakes I made, the parts that surprised me, and the finished product, and creating some fun scenes a long the way.

Appendix

INSPIRATION:

I was deeply inspired by ‘s process and the way he created his animated short through music & visuals. Completely worth checking out:

atom

Every Night I Leave My Body • Short Film & BTS

Yo! welcome to my first post showing a small peek behind the scenes of how I create - hope you enjoy…

Listen now

7 months ago · 255 likes · 37 comments · atom 🐝

RESOURCES + OTHER RELATED WRITING:

Suno Studio Tutorial: A Complete Beginner’s Guide (Youtube)

Create AI Animated Stories & Cinematic AI Videos (Youtube)

Major Copyright Rules Are Changing... Lawyer Explains (Youtube)

I’ve never been good at compressing myself into one role or topic, even when it would’ve made things simpler. I am the intersection of many. And so is my Substack.

But The Off Hours does have a theme. I write about the invisible work – what gets lost, what gets hidden, and what needs protecting when we optimize for speed and scale.

In the coming weeks, I'll give you a behind-the-scenes look at some AI video and animation tools I'm exploring, including the planning and judgment calls you don't always see in the shiny output.

But this week is a personal piece. I can't ask you to understand my work without understanding some of the intersections and contradictions that shape how I see the world. And if you're anything like me – struggling to fit yourself into one box – I hope this gives you permission to stop trying, too.

Rabid Teen

I was a feral, feral teenager. I wasn't scared of much, and I didn't have a great sense of limits.

I remember sitting on the back of my highschool boyfriend's motorcycle as he topped it out on the highway. If you’ve ever gone that fast, your vision looks more like stuttering frozen frames than motion.

I was sneaking out on school nights. Sleeping in grocery parking lots instead of going to class. There were car accidents. My first F. Endless fights with my parents, who couldn’t understand what I was turning into.

It felt like freedom at the time, but I was sad, restless, and confused. Angsty in the way a lot of teenagers are – just very obvious about it.

What’s interesting to look back on is that I hadn’t always been that kid. I was goofy, but I was shy. Highly sensitive. The kind of girl who’d disappear into a book or a craft. Lost hours playing guitar and many choreographed dances with my younger sister.

Somewhere along the way, that version of me cracked open and something less contained came out. But still, music, art, and journaling were the only places I could still make sense of myself. They weren't hobbies. They were how I stayed in tact.

That dramatic shift, from inward to outward, is still part of me. It’s just changed shape over the years.

Everyone Has a Story

The rebelliousness didn't stop, but that energy was redirected.

College was when I felt genuinely inspired by what I was learning. I was paying attention because I wanted to, not because I cared about grades. It was the first time I experienced the feeling of “unlearning” things I had previously absorbed, without questioning them in the first place.

I majored in Journalism with a minor in Sociology. And for years, I would write those degrees off as "useless" before realizing how they’ve shaped me.

I had dreams of becoming a documentary filmmaker: I wanted to tell other people's stories, to travel and meet people far outside my own bubble, and to try to understand how they made sense of the world.

The nagging inner voice wondering if I was doing “something wrong” hadn’t fully formed yet. So I scrappily wrote, narrated and produced my first YouTube “documentary”, before I really knew how to make anything at all.

Meet Ramblin’ Dan, the ‘one-man-band’ who you could find outside the Boston Garden sending “peace waves” to the public.

I come from a family of journalists in Argentina. I'm named after my great uncle, Natalio Botana, who wrote about politics at a time when journalism carried real risk. In that context, journalism wasn't content. It was a responsibility. You asked questions because they mattered, not because they “performed” well.

That history shaped my family in ways I didn’t fully understand until much later. My grandmother (who appears in the first video) and my mom left Buenos Aires because it wasn't safe to be outspoken. And my grandmother was. She spoke her mind openly, publicly challenged authority, and paid a price for that visibility.

I often wonder what she'd think of the world now – where the tools have changed, but the costs of visibility hasn’t. Where women are judged more harshly for using AI and we still feel pressure to prove we're "technical enough."

But every once in a while, you’ll see my “inner journalist” come out here. The part of me that digs deeper than necessary and cares about every source, knowing most people won't even click on them.

I'd be silly not to use AI to help me research and refine, but it doesn't draw the meaning for me. That role is entirely mine. I poke at ideas, cross-check them, and resist flattening something complex into a clean takeaway. I ask what’s at stake and what's possible, and refuse to choose between them.

From Theory to Practice

I often say my life is a web of interconnecting rabbit holes.

I learned how to ask questions through journalism, how systems shape behavior through sociology, and after a transformative dream experience, how memory and emotion construct reality through neuroscience. But those ideas stayed abstract until I took them out of the classroom and into the world.

My dad took a job in the UAE, and I traveled to Abu Dhabi, Dubai and Oman. I had traveled internationally before, but this trip expanded my sense of scale – of who I could become, where I could belong, and how many versions of "normal" existed beyond my own.

When I came home, I landed a job as a designer at a guided tourism company (which is to date, the best job I've ever had). I worked in the marketing department: websites, email campaigns, tradeshow events, light video work for social. And occasionally, I got to travel to the destinations we promoted.

Fast forward a few years: COVID hit, and the industry collapsed.

When the world opened back up, my next chapter was in corporate design: marketing, healthcare, advertising. Bigger systems. Higher stakes. More layers between decisions and consequences. I learned how power moves through organizations – how ideas get shaped, diluted, or abandoned, and how “efficiency” can sometimes act as the stand-in for judgment.

And then, of course, AI became part of the conversation.

I was genuinely excited and curious. I took a course, built internal experiments, created a “spec project” to test what these tools could actually do. I wasn't interested in whether AI could generate something quickly – I wanted to know whether it could support better thinking, and eliminate the day-to-day pain points that designers often face.

Burkett's Bees Presentation Final

69.4MB ∙ PDF file

Download

AI Tools – Spec Work

Download

That curiosity turned into an Art Director AI Integration role. On paper, it sounded forward-thinking. In practice, it was the turning point.

I started to see what was becoming misunderstood across the industry. The work didn't end at "prompt → generate." That's where the real work began. The more I learned, the less interested I became in hot takes (pro or anti AI) and the more invested I became in asking better questions about what we were enabling.

That's where my perspective on AI really formed. Not from the sidelines, not from theory, but from standing at the intersection of automation and creative work, watching what systems optimize for, and what disappears along the way. And from all the personal detours that came before it.

Which brings me here. To The Off Hours.

So Am I Doing It Wrong?

I spent years thinking the answer was yes. That the deep dives across disciplines were a professional liability. But I’ve realized that the more I integrate these paths, the deeper my perspective becomes.

Apoorvaa Deshpande, a tech lead at Google and an international musician, calls people like us "PolyPaths" – people who pursue multiple paths passionately, not because we can't commit, but because our fullest selves require it. We don't lose focus by doing multiple things. We gain it.

We become more useful, more interesting, more alive.

That’s what The Off Hours is about. It’s a space for people who refuse to be compressed into one label. A place where we examine the invisible work that happens at those intersections – between automation and craft, between systems and humans, between what gets optimized and what gets lost.

It's not about excelling at everything. It’s about being honest about what makes you whole, and building a life around that complexity instead of in spite of it.

Thanks for reading this far. I'd genuinely love to know: what are the 'intersections' in your life that don't fit into a clean bio?

About a lab in Japan that's spent over a decade reconstructing images from the sleeping mind. About how you can fall asleep in a scanner, and on a nearby monitor, the flickering contents of your dreams begin to take shape.

But the more I researched, the more I realized the story isn’t just about dreams anymore.

It’s about the brain-monitoring devices already on nightstands. A billion-dollar government moonshot program. And a future where your inner world – the last truly private space – becomes extractable data.

And that future is closer than I thought.

The Consumer Reality

You don’t have to look far to see the “consumerization” of the sleeping mind. At CES 2026, brain-monitoring devices for sleep were a dominant product category.

Take Prophetic’s Halo: a headband that detects REM sleep and uses focused ultrasound on the prefrontal cortex, claiming to trigger lucid dreaming on demand.

Or the FRENZ Brainband - a wearable that tracks EEG signals and uses bone-conduction technology to "coach" the brain into deep sleep.

None of these devices can read your dreams yet. But they're already extracting emotional states, focus patterns, even your "brain-print" – a neural signature as unique as your fingerprint, but far more revealing.

Legal experts have warned that even “anonymized” neural data can be re-identified. These early devices aren’t just accessories; they are biometric probes building the datasets, algorithms and infrastructure for the next decade.

So it’s not a matter of if dream tech becomes a consumer product.

It’s when – and whether we’ve asked the right questions before it does.

The Lab That Started It All

Much of this story traces back to one researcher: Professor Yuki Kamitani.

Kamitani is widely considered a pioneer of modern brain decoding, having spent decades bridging the gap between raw neural signals and subjective experience.

By 2019, his lab was already using deep neural networks to reconstruct blurry, haunting images from brain activity: shapes and colors that looked like digital ghosts of the dreamer’s thoughts.

Today, they're using generative AI to create “multimodal world models”. Not just what you saw, but the spatial context, emotional tone, and narrative flow of what you experienced.

I should emphasize: this is not a flash-in-the-pan experiment. Between elite government grants and deep international ties to prestigious institutions, Kamitani’s lab has the financial runway to scale his tech into global infrastructure.

His work is also the cornerstone of Japan’s billion-dollar Moonshot R&D Program – a national initiative whose stated goal is:

Read that again. The program’s roadmap explicitly aims for a future where you are no longer trapped by the limits of your physical body. Or time. By 2050.

This vision is anchored in the development of the Internet of Brains (IoB), a project designed to connect human consciousness directly into digital networks. By 2030, the government's goal is for one person to operate 10 or more cybernetic avatars simultaneously via “brain-link” – a feat that requires a massive, robust infrastructure, not just a lab setup.

Once governments invest at this scale, you know the commercial sector and every other world power is eyeing it from the sidelines.

And while Kamitani’s lab builds the highway, others are building the vehicles.

• UC Berkeley and UCSF built a device that streams thoughts into audible speech within one second, and can learn new vocabulary it was never trained on.

• NTT's 'Mind Captioning' technology generates text from recalled memories, even when certain language regions in the brain are inactive.

• And at Fudan University in China, researchers proposed a new framework to decode dreams into coherent video – translating sleep imagery, frame by frame.

Researchers are no longer just observing how the mind works; they are capturing, extracting, and reconstructing thought itself before it ever reaches a muscle.

Kamitani is deeply committed to transparency and helps shape UNESCO's global neuroethics framework. But in a race this competitive, it's impossible to guarantee everyone shares the same values.

The “Neuroverse”

It's hard not to admire the sheer ambition behind Kamitani's work.

Much of the research is motivated by genuinely humane goals: helping psychiatric patients visualize hallucinations for therapy, allowing paralyzed patients to communicate through imagery, revealing how trauma is processed during sleep.

But Kamitani's ultimate vision goes far beyond the medical.

He calls it the ‘Neuroverse’, and imagines a future where:

• Dreams become digital files you can record, replay, and share

• Art materializes from pure thought, no physical creation required

• Memories transform into spaces others can explore – invite your friend or doctor!

To get there, researchers are racing towards something called the “zero-shot neural decoding” – systems that can translate anyone's brain signals without lengthy personalization.

Right now, your brain is still unique enough that a decoder trained on someone else won't work on you. That’s a natural privacy barrier.

But if that barrier falls, an uncomfortable truth comes into sharp focus:

Dreams are data. And in our world, data is rarely left alone.

What's at Stake

Once dreams become data and data becomes accessible, the question is no longer what’s possible – but who gets to use it, and for what purpose.

The following three examples are no longer abstract concerns.

1. The “Dream Ad” Problem

In a 2021 survey by the American Marketing Association, 77% of marketers at 400 different firms said they planned to deploy “dream-tech” for advertising within the decade. And we’ve already seen the first move.

That same year, Molson Coors partnered with a Harvard researcher on a Targeted Dream Incubation (TDI) experiment. Participants like pop star Zayn Malik of One Direction watched videos of waterfalls and beer cans, then slept to soundscapes designed to plant a "refreshing Coors Light dream." The experiment worked. And Coors promoted their “dream ad” throughout Super Bowl LV weekend.

This prompted an urgent open letter from leading neuroscientists at Harvard and MIT, warning that TDI-advertising is not some fun gimmick, but a slippery slope with real consequences.

The takeaway: The desire to hijack your sleep is already a corporate strategy. If an AI can learn to read your internal world, it is only a matter of time before it is paid to rewrite it.

2. Predictive Surveillance

If corporations want to read dreams for advertising, what stops governments from using the same technology for surveillance?

Organizations like UNESCO have already warned about the rise of “neuro-profiling” – using AI to analyze your subconscious data to predict how you’ll vote, shop, or even commit a crime before acting on it.

In China, EEG-embedded caps have been tested on factory workers and rail operators. Now, the government has funded their development for the nation's 500 million drivers to analyze fatigue, stress, and even rage. Alongside self-driving car development, this integrates biological states into the ‘City Brain’ and Public Safety apparatus. It represents the first concerted attempt to industrialize the human 'brain-print.'

The takeaway: Governments and corporations already use data to profile and predict behavior. Neural data would simply extend that framework.

3. Cognitive Liberty and the Courtroom

Could a dream ever be subpoenaed in court? If a specific face or location is identified while you sleep, could that be used as circumstantial evidence?

Today, courts don’t treat dream content as confession. But a field known as Neuroforensics is already emerging – using EEG data to establish “guilty knowledge,” such as whether someone subconsciously recognizes a stimulus they claim never to have seen.

In 2024 and 2025, prosecutors increasingly subpoenaed wearable data to map "emotional spikes" during DUI cases and alleged assaults.

The takeaway: As neuro-wearables go mainstream, your brain waves may be the next category of data on the witness stand.

The Good News

These issues aren’t going unnoticed.

Chile became the first country to adopt neurorights into its constitution. The MIND Act in the U.S. was introduced to address the fact that brain data isn't protected by HIPAA. Organizations like the Neurorights Foundation and researchers like Nita Farahany are pushing for frameworks that treat cognitive liberty as a fundamental human right.

It’s a subject I’m paying much closer attention to. It feels reminiscent of the early social media days, where we all signed the Terms of Service before understanding what we’d give away. Only this time, we may have more of a chance to learn from our past mistakes.

Final Thoughts

While we’re all mesmerized by what AI is doing for us now, we risk losing sight of the technologies developing in the background. The ones that could reshape not just how we work, but who we are.

I don't write this to spark fear, but to shed light on a conversation that's happening but feels hidden by other headlines. Once your internal world becomes readable data, ownership becomes the only question that matters. And it’s easier to protect what’s yours now than to reclaim it after you’ve given it away.

Would you wear a device that let you watch your dreams the next morning? And if you would… who else gets to watch with you?

Do you waste money on AI tools, burn through thousands of credits, and spend hours staring at your computer screen, only to render an image that’s maybe cool but barely usable?

If you answered “yes” to any of those questions, well… you’re in good company. But for the first time, I may have found a design workflow that doesn’t make me want to throw my computer out the window.

Since 2023, I’ve been looking for AI tools I can trust in my professional design work. Tools that can handle strict brand guidelines, offer real editing control, and assist with branding that can scale across websites, decks, social, print, motion.

When AI Meets Girlboss dropped a post about recognizability being the key to standing out on Substack, I watched my entire feed level up their visual identities overnight. I was inspired and eager to get going on mine because instinctively, I knew how true this was.

I dove in headfirst, prompted away… and then, I drowned in options.

I couldn’t commit to anything because I couldn’t recognize myself in any of it. I was ignoring my own advice: The real value isn’t how fast you can generate ideas. It’s how deeply you understand them in connection to what you’re building. Yes, AI will make you faster, but only if you know what you’re creating and why.

What follows is how I got out of my creative rut, found my vision, and built a repeatable system. I’ll show you the exact tools and workflow with video demos inside Recraft.ai, so you can decide if this approach works for you too.

The Diagnostic: When You’re Too Close to See Clearly

I needed the right questions to clarify what I already knew. So I flipped the script and asked ChatGPT to ask me questions.

I'm in a creative rut and need help clarifying a distinctive, sustainable visual identity for my Substack, The Off Hours.

### Role + Approach

Act as a thoughtful brand strategist and creative partner.
Ask me 10 carefully sequenced, non-generic questions designed to surface insight, not reassurance. Avoid clichés, buzzwords, or trend-driven thinking.

### What the questions should uncover

– the emotional core of the brand
– the deeper meaning behind the name The Off Hours
– my values, tensions, and contradictions
– the visual languages I'm drawn to (and actively want to avoid)
– constraints I can realistically sustain long-term

### Goal

The answers will be used to guide visual exploration, prompt writing, and the creation of a recognizable design system.

The goal is to arrive at a visual style that is:
– immediately recognizable  
– sustainable to produce weekly  
– meaningfully tied to how I think and work  
– flexible enough to evolve without losing its identity

Deliver all 10 questions upfront so I can see the full scope and answer thoughtfully.

After answering all 10 questions, I had my constraints: a strict nostalgic color palette, flat design with organic and geometric shapes, and symbolic surrealism through objects.

The Off Hours is about the work that happens outside professional life – the deep thinking, the side projects, the ideas that don’t fit neatly into job descriptions. I wanted a toolkit that could flex across formats: bold illustrations, lifestyle images, background patterns.

If you’re only branding for Substack, you don’t need all this. But the same principles apply: consistency doesn’t have to mean rigidity.

The tools I’m about to show you turned that principle into something repeatable.

The Tools: Discovery vs. Systems

→ If you already have a clear brand vision, jump to Recraft.ai
→ If you're still exploring, start with Lovart.ai

Both platforms share the same high-end engines – Nano Banana, Flux, Kling, just to name a few – but they optimize your brain for different tasks. In my view, one prioritizes discovery; the other prioritizes systems. Depending on how you prefer to work, one might click faster than the other.

Lovart.ai: The Interpretor

I began with Lovart, a mood-first tool that interprets your prompts through a "Mind Chain of Thought" system. You provide emotion and references, and it finds the throughline between them.

Lovart let me stay in my flow. Rather than drag and drop, I enjoyed its click-to-edit feature, where I could easily insert my reference images into any part of the chat.

After dozens of iterations trying to find an illustration style, one remix finally made me pause. It had movement. It was surreal. It was minimal but still kinda chaotic. It had a cookie flying in the background for God's sake. It felt like my room after hours. It finally felt like me.

Lovart is a tool I would utilize in those exploratory stages with a client. To show them what their brand could look like, without investing too much time upfront.

But Lovart is a prototyping tool, not a production system. At the moment, I don’t find it capable enough to enforce consistency or easily scale a brand.

After generating something I could work with, I didn't need more exploration.
I needed a system.

So I headed over to Recraft.

Recraft.ai: The Systematizer

I owe a thanks to Karo, Product with Attitude whose article Refraft 101: A Creator’s Guide to Building a Reusable Visual System introduced me to this tool.

The interface told me this was going to be different, in a good way. While Recraft has a conversational option, it also has sliders, menus, panels for color, artistic levels, style galleries, and vectors you can edit inside the platform. If you're working with a strict color palette or need to scale your brand across multiple formats, here's where Recraft shines compared to others.

Locking In Color, Style, and Precision

Instead of chasing consistency after the fact, Recraft lets you design for it upfront. I rendered a few more images in my new aesthetic, edited them until I was happy, and locked in my nostalgic color palette. From there, my outputs started following the same visual rules.

But here's where it got powerful: there's a slider to reduce the number of colors in your composition. You can then convert your artwork into vectors, giving you a clean file with only a few colors that you can adjust individually – in Recraft or elsewhere.

This is where experimentation gave way to precision, where images started behaving like real design assets.

Here’s a little lay of the land. 👀

Building a Reusable System

Once I had a handful of images that felt right, I turned them into a custom style. Recraft lets you build styles from up to 5 reference images, then fine-tune through testing. It’s a way to stop starting from zero every time.

With that custom style locked in, I could generate entire icon packs, doodle libraries, pattern sets – all maintaining the same visual DNA, all exportable as SVGs. This wasn’t just brand consistency. This is where things turned into a production pipeline.

Permission to Flex

I even tested expanding beyond illustration by applying my style to collage images. There are some refinements I’d make, but overall – same color palette, same visual rules, just a different medium.

The Lesson: You Can't Automate Instinct

The irony is that the more powerful these AI tools become, the more important you become.

You can’t automate taste. You can’t program instinct. The goal isn’t to use more tools –it’s to use the right tools for the right job. And honestly, these tools are in an arms race. Maybe one day there will be a tool that can do it all. (I’m looking at you, Figma x Weavy.ai!)

ChatGPT helped me articulate. Lovart helped me explore. Recraft helped me scale. But none of them told me what The Off Hours should look like. That was still my job.

My system is still being iterated on, tested and is evolving. Build constraints that guide you, not trap you. I didn’t start with tools – I started with a limited palette, a recurring symbol, a flat graphic approach, surrealism through object logic not lighting. Those constraints made everything else possible.

What You Can Do Next

I should mention: I am not sponsored by any of these tools. I'm curious, and I like helping people learn.

If you want to try this yourself:

• I’ve saved the ChatGPT diagnostic as a Custom GPT called “The Brand Therapist”. It helps you define distinctive, sustainable visual identity guidelines, then generates tool-specific prompts so you can actually execute with clarity.

• There are more features to both Lovart.ai and Recraft.ai that I didn’t cover here. Let me know in the comments if you’d like a deeper dive on mockup generators, animation and motion, or typography tools

• Also– if you're wondering how I made those video demos, I used Guidde.com. It's a Chrome extension that records your screen and auto-generates how-to videos with AI. Super easy to edit, and honestly a game-changer for tutorials like this. Get 30% off here.

Drop a comment: What's your biggest creative block right now? Are you stuck on vision, execution, or scaling your work? Do you want to learn more about any specific tool?

I’m here and happy to answer the best that I can.

Happy branding!

Share The Off Hours

My new title had the words “AI” + “Art Director” in it, so I probably should’ve seen it coming. But I’d be lying if I said some the backlash didn’t catch me completely off guard.

To be fair, the majority of the commenters were supportive, encouraging, and genuinely insightful. But of course, the negative ones stuck.

I was called a thief. Accused of not caring about the ecological damage AI wreaked. I was “planning to participate in the active destruction of communities and hurt people.” Someone even attacked my grammar (which, ironically, could have been improved with the use of AI.)

But I don’t share this story for pity, to call these people out, or to find an echo chamber that will just agree with everything I say.

I share it because seven months later, the topic is still nuanced, complicated, and a discussion worth having.

“Slow AI”

Sam Illingworth, author of Slow AI, and his collaborators gave me words I wish I’d had back then. Not in the “damn, I could have showed them!” kind-of-way. But words that could have helped me facilitate better conversations in that LinkedIn thread.

Are AI Images Theft? is a collaborative reflection on plagiarism, influence, and responsibility. It brings together voices from different backgrounds, with different stakes. And it’s absolutely worth a read.

Slow AI

Are AI-generated Images Theft?

This post began in a Note on AI-generated images from Caitlin McColl 🇨🇦, following on from a post about plagiarism in Substack Notes. That earlier conversation was relatively clear. Copying someone else’s words and presenting them as your own is plagiarism. The harm is direct. The boundary is visible…

Read more

8 months ago · 62 likes · 174 comments · Sam Illingworth, Mia Kiraki 🎭, Caitlin McColl 🇨🇦, AI Meets Girlboss, Karen Smiley, ToxSec, Jenny Ouyang, and Karen Spinner

The post acknowledges the legitimate grievances: models trained on copyrighted work without consent, AI replacing commissioned art, the lack of visible sources that makes it impossible to judge influence fairly.

It also addresses the bigger questions: where does influence end and theft begin? And who should be responsible for drawing that line?

What emerges isn’t a necessarily a shared consensus, but their shared principles:

Blame travels upward more effectively than sideways. Transparency builds trust even when disagreement remains. Context matters more than slogans. Ethical use is a practice, not a label.

And the part I loved most …

No one claimed to have it all figured out.

“Steal like an Artist” by Austin Kleon

Many of the perspectives in Sam’s post reminded me of the book Steal Like an Artist by Austin Kleon.

But before anyone gets the wrong idea – this book is not a manifesto for plagiarism.

It’s about how all creative work is built on what came before. How influence is inevitable. The goal isn’t to be 100% original (because that’s impossible), but to absorb what inspires you, remix it through your own sensibility, and make something unmistakably yours.

That’s the difference between stealing “like an artist” versus just plain stealing.

It’s not a perfect comparison – Kleon’s book doesn’t address the thousands of artistic styles scraped without permission.

But the process can feel similar. When I’m iterating on prompts, discerning what works, rearranging outputs until they feel unmistakably mine – there’s something familiar there, in that act of transformation.

The Shift

Outside of my day job, I used to design some of my husband’s band posters. I’ll never forget the moment when he didn’t need my help anymore.

Years of training – grappling with the Adobe’s notoriously tough interface, understanding hierarchy, color, composition, typography, all the design principles – suddenly felt pretty useless.

It became clear that the barrier to entry was lowering, fast. But after the initial sting wore off (pretty quickly), the fact that others could generate beautiful images never really bothered me again.

My husband is creative. He does have an eye.

Plus, why should technical skill always be the barrier between someone and their creative vision anyway?

The Misunderstanding

When people argue that AI is killing craft and process, I hear it. But I think there's a misunderstanding about what “craft” often looks like in a professional design setting.

Some folks assume designers love the entire process, all the time. The pixel pushing. The manual resizing of images for every social platform. The edit request that topples the whole design system like a house of cards. Grinning through clenched teeth: "Sure, we'll take that back to the drawing board.”

Don't get me wrong – I love the craft and always will. But craft was being robbed from us long before AI showed up, and no one was talking about it then. Rushed deadlines. Shrinking budgets. Clients who see us as "the people who make things pretty" instead of strategic problem-solvers.

The real craft happens when I get to slow down. When I can sit with my iPad and draw. Paint something. Take my time with it. But that's not most of my professional work. Most of my work is fast, strategic, and client-driven.

And I can’t say it enough:

Good designers do more than create pretty pictures. 👏👏👏

• A good Designer designs with strategy, taste, and discernment. They understand user needs, brand voice, visual hierarchy, accessibility, and how to solve problems through form.

• A good Art Director directs with intention, creative vision, strategic problem-solving, team guidance, and an eye for brand consistency.

All of that, my friends, AI has not taken away from me. (yet, at least)

Different Use Cases

Dare I say, AI has even sharpened some of my Art Direction skills.

As I practice prompting, I’m exercising a muscle I already have – the ability to direct an image. To be clear about composition. Identify the focal point. Articulate tone and mood. Vague inputs = vague outputs (which is true when directing people, too).

Does AI always interpret my direction correctly? Absolutely not! But sometimes, unexpected outputs lead to different ideas.

🎨 There are other use cases too: Brainstorming. Moodboarding. Creating concepts to test before pitching. Prototyping packaging in different colorways. Generating layout variations to explore directions quickly. Translating design rationale into client-friendly language.

It never ends at a single prompt. For me, the outputs are jumping-off points – ways to nail a client’s vision faster without replacing the actual creative strategy.

Conclusion

When I re-read the comments on LinkedIn, I smell fear. And I get it. I understand why people are uneasy. In many ways, I am too.

Ultimately, those LinkedIn voices made me dig deeper into the subjects that surfaced. The ethical considerations. The ecological damage. The need for authenticity, transparency, and provenance.

No designer wants design to become regurgitated, bland imagery trapped in a smaller and smaller feedback loop. And I genuinely fear the images and videos out there that are blurring the lines even further between what’s real versus fake. It’s legitimately scary.

So are there solutions? Where do we still have a hand, or a say, in all this?

My former colleague and I tried to incorporate some of the ethical topics into our “AI Lunch and Learns” at work. And since then, I’ve been exploring potential solutions, like the blockchain, decentralized platforms, and organizations that are tackling these issue head on.

But as Sam wrote in Are AI-generated Images Theft?:

If we default to accusation, people hide. If we default to indifferences, harm normalises.

We can’t afford for people to shy away from talking about these things. Disagreements? Yes. Hard conversations? Absolutely. But blaming each other and outright rejection? It feels too simple for something this important.

AI often feels like it’s happening to us, not with us or for us.

Which is exactly why I care enough to study it, experiment with it, and talk about it openly.

✧･ﾟ: ✧･ﾟ

I’m doing my best to keep the conversation as human as humanly possible. If you're curious about what I'm learning, the tools I'm using, or if you’re asking the bigger questions too – stick around. We can navigate it all together.

I wanted to give credit again to and all his collaborators who contributed to the Are AI-generated Images Theft? piece: , , , , , , & .

You are all making a big difference in how interact with and think about AI.

1. How to communicate with animals (specifically, my cat Piper) 🐈

2. And what’s really going on in our brains when we dream? ★ ☽

The first is a topic for another time, but the second is the foundation for this post.

We often talk about AI through the lens of automation, workflows, optimization, etc. But as we move toward AGI, I see its biggest potential not just as a worker, but as a translator.

I want it to help us see our blind spots - the gaps between fields of study that don't talk to each other often. Between cultures, species, our inner and outer worlds. I want it to expose what we've been hilariously wrong about all along.

For decades, we called 98% of our DNA "junk" before realizing it was essential. I wonder if we are making the same mistake – treating our dreams like they’re meaningless when it might be the brain's most important work.

We sleep for 1/3 of our lives. Yet our dream worlds remain one of the least explored frontiers of the human mind.

But maybe now is when that changes, as the tools catch up to the questions.

My Dream Journal project will begin as a coffee table book – built from real dreams shared by real people, illustrated and explored across disciplines, beliefs, and fields of study. It will be a place where art, neuroscience, consciousness and cultural storytelling can collide.

The following is a personal story. One that pre-dates my AI work entirely, and explains where this obsession with dreaming actually began.

The Dream That Rearranged Me

In my early twenties, my family was scattering. My mom had relocated to Florida. My dad had lost his job and was about to move overseas. Divorce was on the horizon. Our family home was being sold. The rooms were emptied. The heat was shut off. It was late Fall, and I'd graduated from college months earlier knowing my life was about to look pretty different.

That day was unusually windy. The house felt dark and cold. I took a nap on my old bedroom bed – one of the only pieces of furniture left. My sister’s dog, Aurora, was lying on my chest. I remember her sweet houndy face rising and falling with my breath.

And then I started drifting ～～～～

But wait.

Before I explain the dream, I need to preface: I know how I sound. Kinda nutty, a little delusional, or even like I’m making it up. But I didn’t grow up believing in this kind of thing. And I still don’t fully know what to make of it now.

So I’ll leave it at this: I left my body and floated through my house. I saw younger versions of people I loved. I met someone else. And when, in the dream, I thought my dad was coming home, I snapped back into my body.

I’d never had a dream like that before. And despite years of trying everything – meditation, binaural beats, teas, any technique I could find – I never had one like it again.

But it rearranged something in me. How I thought about reality, memory, the mind, even death. And it sent me down a path I’m still on, because something deep inside me still wants to understand what the hell that was.

The Search

I woke up in that cold room with a desperate need for a vocabulary to describe what just happened. I wrote it all down the best I could. That dream felt more real than my own waking life, and I needed to know whether I was losing my mind or discovering it.

So I went looking for answers. I devoured every book and documentary I could find about the brain, consciousness, and what happens when we sleep. Sure, I stopped at the New Age shelves, but I lived in the neuroscience aisles.

Eventually, I found a book by Rudy Tanzi and Deepak Chopra that seemed to exist in the space between both worlds. It went everywhere with me.

Most of us know Deepak as a spiritual icon, but Tanzi is a titan – a Harvard neuroscientist doing groundbreaking work on the genetics of Alzheimer’s. Their collaboration was the first time I saw two traditionally separate disciplines engage the same topics.

When Tanzi came to Boston for an event, it was sold out and I was ticket-less. So I volunteered as staff to watch his talk from the corner of the room. I waited through the book signing until I was the last person in line. I showed him my dog-eared, annotated copy of his book. We talked a bit about his research and his mother's battle with Alzheimer's.

I didn’t know meeting him would change my life, but in this meandering, uncertain-kind-of-way, it did.

It led me to take a neuroscience course at Harvard Extension School, where, during a lecture on sleep and memory consolidation, I remembered Tanzi’s Mom.

A thought came to me:

Do Alzheimer's patients still dream? And if they could recall them, could it help them regain memory?

I went looking for research for my thesis, but I found a desert. Maybe it was a naïve question. But I was surprised at how little we knew then about dreaming. We understand that sleep is very much tied to memory. We even know that disrupted sleep patterns can be an early indicator of the onset of Alzheimer’s – but the subjective experience of the patient at night? A mystery.

The Craigslist Post

I never got my graduate degree. Life pulled me in all sorts of different directions. But I couldn’t let go of the questions either.

On a whim, I posted a Craigslist ad that simply said:

“Share your dreams with me for an art project.”

I expected it to disappear into the void. Instead, complete strangers wrote back.

Vivid dreams. Mundane ones. Drug-induced ones. Covid dreams. Dreams about clowns. Giraffes. Sleep paralysis. Nightmares. Dreams during grief, after childbirth, lucid dreams, flying dreams, weird, wacky, and all fascinating in their own way.

People trusted me with these personal, half-remembered images.

Even with a small sample size, I started noticing patterns. I took to Instagram, where I began the first attempts at illustrating these stories. I was inspired that people wanted to share their dreams. We are all walking around with these incredible nightly movies with nowhere to show them.

I never deleted those messages; they’ve been sitting in my inbox, like a stack of unopened letters, waiting for me to return to them.

Why Now

What happens when we sleep – our dreams, our brain activity, the entire system – is data. If we treat dreams seriously as data, the question isn’t whether AI can analyze them – it’s what that analysis will reveal about their structure.

I believe dreams are a different mode of reality, constructed and governed by different rules than waking life. Not less real, just differently assembled. If that’s true, then the real question becomes: what system are we actually inside when this happens? ☆

I believe AI is uniquely suited to sit with that question. Not to explain dreams away or over-mystify them, but to notice what repeats, what shifts, and what disappears. To identify dreaming styles and correlate them with memory, learning, creativity, or decline. To surface patterns that only emerge across many minds, over time.

So I’ve decided it’s time to wake this project back up.

The Dream Journal project is a living experiment: part art, part neuroscience, part cultural archive. It starts with a coffee table book, but where it goes from there depends on what we discover.

The goal is to shed light on a subject that’s remained in the dark for too long.

✧･ﾟ: ✧･ﾟ:

This Substack is part of the project now. If you follow along, you’ll get stories, research, artwork, and a front-row seat to building this thing.

But I'm also looking for collaborators. If you work in sleep science, neuroscience, AI research, DeSci, or anything touching dreams and consciousness – or if you're just deeply curious – I want to talk to you. And I'm still collecting dream submissions. ☺

“AI makes you dumber.” “AI is ruining human creativity.” Or even better – “Why don’t you just run it through AI?”

None of them capture what’s really going on.

Ironically, my friends who reject it are the most likely to send me an AI generated piece of content without a clue. And the ones who swear by it act like the work ends as soon as you hit enter.

It’s been five months since I left my 9-5 and finally, my thoughts on AI are starting to take the shape of full sentences.

But first, a bit of context on how I got here.

It was October 2023, the dawn of mainstream AI

That year, I had taken a monthlong sabbatical from work. During that time, I got married, started a passion project, and set out to learn as much as I possibly could during that rare time to myself.

My eyes were set on an animation course – but then, a phone call from my Mom.

“People at work are talking about this course on AI. Maybe it’s worth looking into.”

At the time, my Mom was producing video for Coindesk, a leading media company focused on cryptocurrency, blockchain and the Web3 space. She didn’t land in tech through a master plan or a finance degree. She spent years attending free MIT forums, networking, and doing her own research long before most people knew what crypto even was.

Ever since I can remember, she’s been fascinated by new technologies that quietly shape how we work, live and participate in the world.

I often think about how what we say about AI now will sound so funny in the future.

To illustrate, here’s a shortened clip from a video my Mom produced in the 90s called Email for Everyone.

One of my favorite lines:

“To a lot of users, the power of email is getting a letter or message any time of the day, any day of the week, including Sundays and government holidays, even when you’re traveling!”

Oh, how times have changed.

Public libraries put it on their shelves. PBS stations aired it. Orders came in by fax – pages piled up on her bedroom floor.

We laugh about it now. The script, the pacing, the whole thing feels like a time capsule. But for a moment, email was something you had to explain. Something that required a whole video just to make it feel approachable.

Now, it’s so woven into our lives that we forget it was ever new at all.

So when my Mom gently suggested I learn about AI, I knew she wasn’t chasing a trend. She was sensing a shift.

I took the course.

Returning to work with bright-eyed enthusiasm

I blinked and I was back at my job.

After the sabbatical, I felt energized. Excited to show my colleagues how AI could handle the tedious stuff – then we could focus on the meaningful work!

Outside work hours, I created a full ‘spec brand’ to test what tools could actually help us at our Creative Agency. I explored every tool I could get a free subscription to: presentation generators, mockup software, a custom GPT trained to speak in our voice.

I wanted to see where AI could genuinely expand creative possibilities and where it hit a wall.

The project (below) became my sandbox.

Burkett's Bees Presentation Final

69.4MB ∙ PDF file

Download

Download

I presented my findings to my team, and in hindsight it feels like a student bringing in a homemade science experiment to share.

My enthusiasm at work turned into an unofficial title: “AI Girl”. Then it became an official title: Art Director, AI Integration. It sounded exciting but I didn’t have the time, authority, or honestly, the knowledge to implement anything meaningful.

Plus, who did I think I was trying to “make AI work”?

Meanwhile, the ethical questions got louder. Energy concerns. People worried about losing jobs. AI slop. Deepfakes. The more I learned, the more complicated it became.

My enthusiasm softened into something much more cautious. A kind of quiet questioning of what I was really enabling.

So how do I feel about AI right now?

The short answer is: conflicted, curious, and trying to ask the right questions.

AI is holding up a mirror to our species. It’s forcing us to ask what makes us human – and what we want to protect.

Geoffrey Hinton, the “Godfather of AI,” famously put it this way:

“The best way to understand it emotionally is we are like somebody who has this really cute tiger cub. Unless you can be very sure that it’s not gonna want to kill you when it’s grown up, you should worry.”

So if AI is still in its infancy, I want to understand how it’s growing, who’s feeding it, and whether we still have a chance to tame it before it’s too late.

Here’s what I’ve been thinking about…

Intention matters more than ever

Protecting craft and values.

I respect the leaders and creatives taking a stance.

When asked about generative AI, Filmmaker Guillermo del Toro said, “I’d rather die” than use it.

He goes on to say his concern is not artificial intelligence itself, but natural stupidity. Mindless dependency. How it’s being used, not that it’s being used. He set a boundary: a refusal to automate away the craft that makes his work meaningful.

Other brands are naming what they want to protect. Dove pledged not to use AI-generated women in its campaigns. Polaroid’s creative director talks about protecting the “imperfect, analog layer” in all of us. Some Fortune 500 brands have added clauses to agency contracts: AI isn’t banned, but it can’t be used thoughtlessly.

These aren’t necessarily anti-AI stances, they’re value stances. And the market can feel the difference. Even beautifully crafted AI ads still trigger more distrust in testing.

Our brains pick up when something human has gone missing.

(For now, at least.)

Seeing is no longer believing

Protecting verification and provenance.

We’re approaching a moment when photos and videos no longer function as proof. That’s destabilizing in ways we’re only beginning to understand.

My instincts have already shifted: I second-guess content I wouldn’t have questioned a year ago. I go straight to the comments. I cross-reference. But I know those small efforts will be futile soon.

Brands and creators have always focused on crafting a compelling story. Soon they’ll have to focus on protecting it too. When anyone can generate convincing content in seconds, the question isn’t just “does this resonate?” – it’s “can we verify this is ours?… or prove when it isn’t?”

When I mention the “blockchain,” I may lose some readers. It’s been associated only with the finance bros, crypto-savvy and in my opinion, misunderstood by many.

But at its core, it’s a ledger. A record. Designed to be transparent, immutable, and verified by a network.

It’s already being used in interesting ways.

• Louis Vuitton, Prada, and Cartier formed a consortium to verify product authenticity on their own Blockchain – you can scan a handbag and trace its entire life cycle back to the source.

• Musicians are selling directly to fans on Ethereum platforms, bypassing Spotify’s fractions-of-a-cent royalties.

• A new layer of the internet is emerging where authorship can be proven, edits aren’t invisible, and creators actually get paid.

These systems aren’t perfect or fully mature. But they point toward a world where provenance and credit aren’t afterthoughts – they’re built into the foundation.

We still get to decide how power is distributed

Protecting ownership and data.

Most of the AI we use runs on centralized infrastructure. A handful of companies control the data, the training, and the outputs. They decide what the models learn, what it refuses to say, and what gets prioritized.

But my real fear: mix AI with purely centralized systems and you’ve combined something extremely efficient with something extremely opaque.

These become systems you can’t audit or opt out of.

It sounds like science fiction until you look at China’s social credit system, algorithmic hiring tools that reject candidates without explanation, or personal accounts getting shadow-banned with no way to appeal.

That’s why I’ve started paying attention to the decentralized systems being built alongside it.

The core idea isn’t new – Satoshi understood it when he built Bitcoin. Instead of routing everything through a single authority, you spread it across independent nodes, coordinated by transparent rules.

Decentralized AI applies that same logic: your data, your prompts, your creative work doesn’t have to flow through one company’s servers. It turns AI from a product you rent into an ecosystem you can participate in, shape, and even co-own.

As of 2025, this space is no longer purely experimental.

• Render Network lets creators access GPU power from a decentralized network instead of renting it from Amazon or Google.

• Bittensor rewards contributors for training AI models that compete with each other.

• The Artificial Superintelligence Alliance is building a decentralized marketplace for AI services – a counterweight to the “one model to rule them all” approach.

The more I learn about decentralization, the more I realize: protecting ownership and data might be the only way to protect the ability to choose at all.

The human layer still matters

Protecting slowness.

There are many themes worth protecting that I’ll save for a different Substack, but I want to sign off with this one:

If AI is the fast layer – the accelerant, the compressor, the thing that distills and automates and moves quicker than any of us can think – then it feels more important than ever to protect slowness.

Not inefficiency. Not laziness.
Just intentionally making room for the human pace.

Across my industry, I watched timelines shrink. Expectations shift. Misunderstandings surface about what AI can and cannot do. For many of us, our jobs didn’t change, but the perception of the work did.

Businesses know that AI is already in the room. It’s in your client’s expectations. It’s in your competitor’s workflow. But if creative thinking exists in all business – which I believe it does – true creativity still depends on human time, patience, and discernment.

Even prompting takes intention. Iteration takes time. The outputs are often only as good as the inputs. And occasionally, the “AI shortcut” is longer than the old-fashioned way.

But AI hasn’t killed my desire to create without it. If anything, it’s pushed me toward the slow stuff I can’t automate away. Picking up my guitar. Writing in my notebook. Reading, painting, or going for a walk. Exercising all the muscles that make me feel human.

So in this moment when the ground is shifting and we have to decide what still feels sacred, I’m choosing to slow down long enough to ask.

Conclusion

Before I know it, this essay could read like Email for Everyone – outdated, obvious, maybe a little silly. But that’s kind of the point.

My Mom made that video not because she had all the answers, but because she wanted to help people understand what was happening while it was taking place. The choices they had. The new territory they were stepping into.

I’m trying to do the same.

I should probably start with a confession: I didn’t leave my job because things were terrible.

In fact, it was the opposite. I was getting promoted, getting face time with leaders I genuinely admired, and finally proving to myself that I could survive in a corporate environment without spontaneously combusting.

But sometimes the place that teaches you you’re capable is the same place that shows you you’ve outgrown the role.

For me, the shift wasn’t dramatic. No meltdown, no flaming Slack messages, no anger. Just a quiet internal click – the sense that the world outside my job was accelerating faster than what was happening inside it.

Everywhere I looked, entire systems were being rewritten. Blockchain reinventing finance and trust. AI reshaping how we work and create. Science, art, healthcare – all of it evolving at a pace that felt impossible to ignore.

And then it dawned on me: If the world is reinventing itself, why can’t I?

The Treadmill

Right before I left, I’d been promoted to Art Director, AI Integration. I was grateful. For a moment, I had time built into my job to explore something that felt important – to research, experiment, help others understand what was unfolding.

But for the majority of companies, the treadmill never stops. Work moves. Deadlines tighten. Clients demand. AI quietly becomes a side project tacked onto a full workload.

Real integration takes infrastructure – training, experimentation, room to fail. Most companies aren’t set up for that yet. Not because they don’t want to be, but because the path forward is still murky: security concerns, budget questions, workflows and processes that don’t even exist yet.

Everything feels uncertain. For everyone. Maybe it’s that same uncertainty that pulled me in — the sense that something larger is shifting beneath us, and I couldn’t afford to ignore it.

So I jumped.

Many Threads

What I’ve realized in the three months since leaving is that quitting a job doesn’t necessarily reveal your next chapter. It just removes the noise long enough to hear your own questions.

Aside from asking myself if I’ve ruined my life, I keep circling back to the same thought: Who am I, what am I creating – and who am I building it for?

I can’t fully answer that yet. But I’ve been pulling on many threads: producing video for a blockchain startup, helping companies figure out what AI actually means for them, and picking up projects that look unrelated on the surface but keep teaching me something new.

I’ve never fit neatly into one category. Even my education was a patchwork: journalism, sociology, neuroscience, design. For years I treated that mix like a professional liability. Everyone else had a lane. I had a map of intersecting rabbit holes.

The Complete Human

Then I saw a post from Steven Bartlett.

He said the specialist may not survive AI – the people who will thrive are “complete humans.” The ones who build and create and think and feel across disciplines.

What he wrote reframed everything I’d been treating as a flaw. The painting. The guitar. The deep dives in neuroscience that never connected to a career path. Maybe those aren’t distractions from the work. Maybe they are the work.

And there’s an irony I keep turning over: what if AI – an artificial thing – is part of what helps us become more human? Not by replacing us, but by forcing the question of what was worth protecting in the first place.

One thing I need to make clear: I don’t learn about AI because I love AI. Truthfully, it scares the sh*t out of me for a variety of reasons.

I learn about it the way you’d study weather patterns on the coast – not out of love for the storm itself, but because it’s coming either way.

How will it take shape? What’s the hype? What’s the human cost?

Those are the threads I’m pulling on.

What This Is

As a private person, writing publicly feels strange. But this isn’t performance. It’s processing. Thinking out loud while the ground shifts.

If you stick around, you’ll find me exploring these questions from many angles: tools reshaping creative work, the promises and failures of emerging tech, a curious generalist trying to make sense of a world moving faster than any single discipline can explain.

I don’t have answers. But I think the questions are there.

See you soon.