At Hustle Fund, we just announced that Dunky AI, our cute, lovable, purple hippo-esque friend has taken our the company and has gone rogue today on April 1, 2025. Hopefully we can wrestle back the company tomorrow!
In the meantime, Dunky AI has chosen to write a strange comic book, set in a mythical place called Hippoland…a world that exists to teach entrepreneurial education. Talk about hallucinations!
While this is all in fun for the day, I’m going to give you a serious behind-the-scenes look at the creation of said comic book and some learnings. This project ended up being a ton of fun and didn’t take long.
If you have imagination, you can be an artist and creator — even if you have zero talent
This is a bold statement and was not possible prior to the last couple of years. I started this project in early 2024 and put very little time into this throughout the year – probably a total of 1-2 full working weeks. But, throughout that year, the technology has only gotten better. This comic book is a reflection of last year’s tech.
Midjourney generated almost all of the images. Adobe Photoshop’s AI tools aided as well. Cartoon people in this book were created based on seeding these AI tools with real photos. In some cases the cartoon representations of my team members were fairly on-target:
These drawings are based on real photos of my colleagues Haley Bryant and Kera DeMars and arguably look pretty similar to their actual likeness.
But, Midjourney didn’t get it right for others on our team, so precision in drawing has a lot to be desired still. I think a lot of generative AI software is good at drawing generic people but not specific people you may have in mind.
In addition, this cartoon of me is based on a real photo of me wearing a grey shirt that reads Hustle Fund. You can see Midjourney takes some artistic liberties in changing the shirt 🙂 :
Generative AI is good at drawing people and bad at drawing hippocorns
This may seem like a silly learning, but extrapolated, this is a big shortcoming of generative AI tools these days. If you want to draw generic people, there are MANY tools that can do this well. If you want to draw a mythical creature that doesn’t exist in the exact form that you are envisioning, this is near impossible. This makes sense, because there is limited to no data for these models to train on. Afterall, as much as my inner voices in my head may disagree, hippocorns do not exist.
But, when you think about blockbuster characters that have done well over history, they are ALL mythical. They are Mickey Mouse and Elmo. They are Pokemon and Yoshi. These characters are all figments of people’s imaginations, and it is really hard to go from your head to paper.
I struggled to get our hippocorn to be close enough to what we envisioned. You can see that it’s not quite right. Our mascot Dunky is not as wide and is not as old. It has buck teeth, which AI often misinterprets as a shirt or an underbelly of white. I did the best that I could in cartoonifying our plushie.
Consistent characters are still a problem
Midjourney (and others) launched a consistent character feature in their software during the creation of this comic book. But, this feature is not actually that accurate, especially when it comes to specific objects (especially made up creatures like hippocorns). This is probably the top problem that still needs solving in generative AI. Once this is solved, we should see independent Disney-competitor brands started by 1-person design shops (who may not even be artists)!
Much like how we now have vibe coding and 25% of YC companies in the latest cohort have codebases largely written by AI, we will see the same for design and art. Along the way in doing this project, I met a kid who is in a university but spends almost all of his time consulting for companies who want to generate images and need his help with AI prompting. I don’t think this skill or job stays forever, as it will become easier and easier to prompt images and videos without a consultant, but I do think it represents a shift in a how people will do design. It won’t be by hand anymore. Maybe it will be 80% by AI and 20% by hand (editing the parts that need help that AI can’t get right).
When Disney came out with the original Snow White and the Seven Dwarfs, it cost $1m+, which was a crazy amount of money back then! They maintained edge in this world of character creation, because they established a large capital moat and could employ the best artists and storytellers to create hit after hit. Even when they didn’t hit, it was ok, because they had more shots on goal coming soon. Even when they ran into a wall and ended up buying Pixar to invigorate the brand, stories, and technology, money was the moat. Very few studios could even afford that.
Now, there is no moat for anyone, because the cost to create blockbuster movies and cartoons is so close to dropping to near zero cost. I’m excited to see the next generation of characters and worlds that will influence the world.