Ankle to Apps: AI Journey

With Shannon Cascio – I just got recognized as one of their top fans! 🎉

Yesterday I entered a Lovable contest with the invoice
app I built from my couch.

Sounds small. Felt huge.

For the past few days, I've been refining InvoiceCraft,
my no-login, no-subscription invoice tool for
freelancers and yesterday I finally submitted it.

The hardest part wasn't building the app. It was
writing the description. They asked: "What are you
building?" and gave me a tiny 200-character limit.
TWO HUNDRED.

I rewrote that paragraph 14 times. Fourteen.

Every time I started writing, I sounded like a
corporate brochure. "Innovative solution leveraging
cutting-edge technology to revolutionize..." 🙄 No
one cares. I had to keep cutting until what was
left was just... me. A person recovering from
ankle surgery who built something that actually
helps freelancers.

I have no idea if I'll win this contest. But for
the first time in three months, I submitted my
work somewhere and meant it.

That's the win.🏆🛋️

If you'd like to see the video demo comment "DEMO" below

DAY 02 / 90 🐼⚡
The fourth test today was music generation using [Suno](https://suno.com/)

After building:
🎭 the ad scripts
🎙️ the voiceovers
🖼️ the images
🎬 the videos

…I wanted to see if the same AI workflow could also generate actual songs.

So I opened Suno and started experimenting.

The process was surprisingly simple:

• Open Suno
• Click “Create”
• Paste in the generated lyrics
• Paste in the music style and mood prompts
• Give the song a title
• Click Create
• Then sit there slightly confused while AI starts making music from your idea

The weird part wasn’t that it worked.

The weird part was how emotionally convincing some of the songs actually sounded.

Some came out hilarious.
Some sounded like real commercial jingles.
Some sounded better than songs I’ve heard in actual ads.

And once again, the same pattern kept showing up:

The quality of the result depended heavily on the quality of the direction.

Tiny changes in:
• genre
• tempo
• vocal style
• mood
• instrumentation
• lyrical pacing

…completely changed the outcome.

That’s becoming the biggest lesson of this entire journey:

AI is not replacing creativity.

It’s amplifying direction.

The people getting amazing results aren’t just clicking buttons.

They’re learning how to think like:
🎬 directors
🎵 producers
😂 comedians
🧠 storytellers
📣 marketers

All at the same time.

And now the pipeline is starting to feel complete:

PunchlineAds creates the concept.
ChatGPT expands the scripts and prompts.
ElevenLabs creates the voiceovers.
Google AI Studio creates the visuals.
Google Flow creates the videos.
Suno creates the soundtrack.

One funny business idea can now evolve into an entire multimedia campaign using AI tools.

A week ago this all felt overwhelming.

Now it’s starting to feel possible.

Still messy.
Still experimental.
Still learning constantly.

But every day the pieces connect a little more.

Suno | AI Music Generator Create stunning original music for free in seconds using our AI generator. Make your own masterpieces, share with friends, and discover music from artists worldwide.

DAY 02 / 90 🐼⚡
The third test today was video generation using [Google Flow](https://labs.google/flow/tv)

And this was the moment where everything started feeling less like “AI experiments” and more like an actual creative production pipeline.

After generating:
🎭 the ad script
🎙️ the voice
🖼️ the images

…I wanted to see if the same prompts could evolve one step further into actual video scenes.

Here’s the workflow:

• Open Google Flow
• Click “Create with Flow”
• Start a New Project
• Paste the same visual prompt
• Hit Send
• Wait for the AI chaos to begin 😂

A few days ago these ideas only existed in my head.

Now they’re turning into moving scenes with characters, camera movement, environments, and cinematic shots generated by AI.

But something important became obvious very quickly:

Just because AI CAN generate video doesn’t mean the first result is good.

Some outputs looked incredible.
Some looked completely broken.
Some accidentally became funnier than the original joke because the AI interpreted things in weird ways.

That’s another thing people don’t talk about enough:

A huge part of using AI creatively is learning how to guide randomness instead of fighting it.

The better your prompts become:
• the better the pacing
• the better the camera angles
• the better the expressions
• the better the storytelling

You stop thinking like a “user” and start thinking like a director.

And now I’m starting to see the bigger picture of what I’m actually building here:

PunchlineAds creates the joke.
ChatGPT expands the concept.
ElevenLabs gives it a voice.
Google AI Studio creates the visuals.
Google Flow turns it into video.

One idea now moves through an entire AI-powered content pipeline.

That’s honestly blowing my mind a little.

Still learning.
Still experimenting.
Still breaking prompts daily.

But for the first time, I can actually see how a solo creator could build an entire content studio with AI tools.

DAY 02 / 90 🐼⚡
The second test today was experimenting with image generation inside [Google AI Studio](https://aistudio.google.com/)

After generating the ad scripts and voice content, I wanted to see if I could create matching visuals for the ads without touching Photoshop or spending hours designing.

Here’s the workflow I used:

• Open Google AI Studio
• Go to Playground → Models → Image Generation
• Turn OFF “Grounding with Google Search”
• Select the Nano Banana model
• Enter the prompt
• Watch the madness begin 😂

AI image generation is less about “creating art” and more about learning how to direct a scene.

You’re basically becoming:
🎬 director
📸 camera operator
🧠 creative strategist
😂 comedy writer

All through prompts.

And the craziest part?

This entire workflow started with one generated joke inside PunchlineAds.

Now that same joke can become:
🎭 an ad script
🎙️ a voiceover
🖼️ an image
🎵 a song
🎬 eventually a full video ad

That’s when you start realizing these AI tools aren’t isolated tools anymore.

They’re connected creative systems.

Still early.
Still learning.
Still breaking things constantly.

But this is starting to feel very real now.

DAY 02 / 90 🐼⚡
One small thing I learned today while testing the PunchlineAds workflow:

AI voice generation is way more than just “paste text and click play.”

The first test was in ElevenLabs (elevenlabs.io) using the Rachel voice.

At first… honestly?
It wasn’t funny at all.

The script itself was decent, but the delivery sounded flat and robotic. The jokes had no timing, no personality, no energy.

Then I noticed the “Enhance” button.

I clicked it, and ElevenLabs automatically added emotional and performance-style tags throughout the script.

Suddenly the same exact words sounded completely different.

Rachel started sounding more expressive, more sarcastic, more human.

And that’s when something clicked for me:

In comedy, delivery matters just as much as the writing.

The joke alone isn’t enough.
Timing, pauses, emphasis, emotion, and pacing completely change how people experience the content.

That was a bigger lesson than I expected today.

AI tools still need direction.
They need shaping.
They need iteration.

The people getting great results with AI usually aren’t using better tools.

They’re spending more time refining the output.

Tiny adjustments can completely change the final result.

It’s honestly fascinating watching the pipeline slowly come to life.

🎭 Script → 🎙️ Voice → 🎵 Music → 🎬 Video

Tomorrow I’m continuing the experiments and seeing how far I can push the humor and personality in the generated content.

DAY 02 / 90 🐼⚡

Today was the first real “builder day.”

No more planning.
No more researching.
No more asking AI for ideas.

Today I actually built something.

I created the first version of my micro-app:
⚡️PunchlineAds

The idea behind it is simple:
Generate funny ad concepts and scripts for businesses using AI.

What sounds simple on paper turned into several hours of debugging, testing prompts, fixing broken outputs, and trying not to burn through Lovable credits too fast.

Here’s what I actually did today:

• Used Claude to generate the prompt architecture for the app
• Took those prompts and manually entered them into Lovable one by one using Plan Mode
• After EVERY prompt, I tested the app for bugs and weird behavior before moving forward
• Repeated that cycle over and over for hours

By the end of the session:
✅ First working version built
✅ Multiple prompts tested
✅ Workflow validated
❌ About 50 Lovable credits destroyed in the process 😂

One thing I learned VERY quickly:

AI-generated apps are not “push button magic.”

The internet makes it look like people type one sentence and suddenly have a million-dollar SaaS.

That’s not reality.

The real process looks more like:
Prompt → Test → Break → Fix → Retest → Confusion → More fixing → Tiny breakthrough → Repeat.

But honestly?
That’s where the learning happens.

The coolest part came AFTER the app was working.

I took the generated PunchlineAds output and fed it into ChatGPT to create:
🎙️ ElevenLabs voice prompts
🎵 Suno song prompts
🎬 Google AI Studio video prompts

That’s when it clicked for me:

I’m not just building random apps.

I’m building a connected AI content creation pipeline.

One tool feeds the next.
One output becomes another input.

That realization completely changed how I’m thinking about these projects now.

🚀 Tomorrow’s Focus:
Refine the app outputs and start generating actual funny ad content and videos with the pipeline.

👇 If you want to test the PunchlineAds app for yourself, comment:
PUNCHLINE

I’ll send you the link.

DAY 01 / 90 🐼💻

Pain Level: 1-2

A few days ago, I injured my left ankle badly enough that surgery is now part of the conversation.

Which means for the next 90 days, my life is going to slow down whether I like it or not.

At first, I was frustrated thinking about everything I wouldn’t be able to do.

Then I realized something:

I can either spend the next three months feeling stuck… or I can use this time to completely reinvent myself.

So today I started building a 90-day plan.

Not just random goals, actual projects and skills that build on each other and lead to something real by the end.

Here’s what I want to learn while recovering:

• Build micro-apps using tools like Claude, Lovable, Replit, and AI-assisted coding
• Create AI-generated branding, videos, and visuals with Gemini and AI Studio
• Learn a few songs on keyboard and guitar
• Draw 30 kawaii/chibi characters just for fun and creativity
• Improve my pencil drawing skills by practicing portraits and faces
• Most importantly: document the entire process publicly

Today I spent hours talking with ChatGPT, Claude, Gemini, Manus, and Perplexity trying to figure out what’s realistic and what’s just me getting distracted by shiny ideas.

One thing hit me hard:

AI gives you infinite possibilities. That’s dangerous if you don’t know how to focus.

Some of the tools pushed me to think bigger.
Some pushed me to simplify.
ChatGPT basically told me I was trying to do too much in 90 days.

I have a custom instruction in ChatGPT that makes it brutally honest. If you'd like those instructions, comment HONEST below.

And honestly… it was probably right.

The biggest lesson from Day 1:
Using multiple AI tools is powerful, but it can also become mental chaos if you keep chasing more ideas instead of executing one.

So tonight I combined the best ideas into a single plan and committed to moving forward.

Tomorrow I stop planning.

Tomorrow I start building.

🚀 Day 02 goal:
Start creating the first micro-app.

Received some encouragement and inspiration from my AI guru friend Shannon Cascio. Thanks, you're the best.