The Digital Toybox: Exploring the Frontiers of AI-Generated Collectibles

By Mike | Last Updated: June 22, 2025

In the rapidly evolving landscape of artificial intelligence, the boundaries between reality, nostalgia, and digital hallucination are becoming increasingly blurred. What began as a creative pastime following a 2025 "Flower Power Cruise" has transformed into an intensive study on the capabilities and quirks of advanced generative AI models. As I continue to design custom "action figures" for friends, colleagues, and cultural icons, the process has revealed not only the incredible potential of prompt engineering but also the persistent, often humorous, challenges of machine “hallucinations.”

The Genesis of a Digital Hobby

The project was born out of the 7-day Flower Power Cruise, which departed from Fort Lauderdale last March. The voyage was a nostalgic journey through the 1960s and 70s, featuring legendary musical acts such as The Cowsills, Jay and the Americans, Ron Dante, and Jefferson Starship. With stops in Falmouth, Jamaica, and the Cayman Islands, the cruise provided a treasure trove of memories and interpersonal connections that I felt compelled to immortalize in a unique way: the creation of custom-carded action figures.

Using ChatGPT as my primary design engine, I began synthesizing personal photos, headshots, and historical artifacts into tangible-looking digital collectibles. However, as any frequent user of OpenAI’s o3 and o4-mini models can attest, the path to the perfect digital asset is rarely straightforward.

Navigating the AI Hallucination Labyrinth

One of the most critical aspects of working with current generative models is understanding the phenomenon of "hallucinations." In the AI world, this occurs when a model produces information—or, in this case, visual elements—that deviate from the user’s intent based on internal training data or erratic logic.

According to internal company benchmarks, OpenAI’s latest models, despite their sophistication, still exhibit hallucination rates of 30-50%. These glitches manifest in various ways: the sudden appearance of multiple "Groovy Bears," duplicate milkshakes, or the unexpected transformation of a soda shop setting into the cinematic aesthetic of the film Grease. For the creator, this requires a delicate balance of patience, iterative prompting, and a willingness to accept "happy accidents" when the AI decides to inject its own creative flair.

Chronology: The Evolution of Design

The development of these figures has been a process of trial and error. My first attempt, "Martha," serves as the perfect case study.

The Martha Project: A Lesson in Iteration

My initial attempt to create a figure of my friend Martha involved a classic poodle skirt and white blouse. The AI, however, insisted on populating the scene with two Groovy Bears and two chocolate milkshakes. Subsequent iterations saw the inclusion of the band "The Archies"—likely because the AI associated the setting with Pop’s Soda Shoppe—and even a complete aesthetic shift toward a 1950s cinematic look.

It was only through rigorous, iterative tweaking that I was able to achieve a final, proportional version of the figure. This experience highlighted a recurring issue: the more one regenerates a figure, the more the output risks sliding into a "Toon" aesthetic, reminiscent of the characters in Who Framed Roger Rabbit.

Capturing Life’s Artifacts: The Peter Noone Experience

One of the most successful projects involved Peter Noone of Herman’s Hermits. During the cruise, Noone famously engaged with the audience, throwing a fan’s purse and badge onto the stage and teasing a man in gold sparkle shoes. By integrating these specific, real-life artifacts—the purse, the badge, and the shoes—into the action figure’s design, I created a piece of memorabilia that felt rooted in the shared experience of the cruise.

Supporting Data: From Professional Branding to Personal Keepsakes

The utility of this technology extends beyond nostalgia. I have utilized these techniques in professional settings, creating figures featuring public figures like Rob Gronkowski and internal company leaders. These experiments were designed to test the limits of branding and likeness accuracy.

Archeology and Nostalgia

The project regarding Dr. Neil V. Hitch remains a highlight of the collection. Dr. Hitch presented findings from an archeological dig at the site of the 1969 Woodstock festival. By using his headshot, images of the recovered items (such as vintage soda cans and bread wrappers), and iconic imagery from the original festival, I constructed a figure that served as a historical narrative on a single card.

The "Purge" and Other Creative Departures

My colleague, Eric Carlson, took the experiment to a new level by applying a The Purge theme to a figure of our director. This highlighted how AI can be "steered" into specific, albeit darker, creative directions when the prompt is sufficiently descriptive. Other projects, such as the figure of my grandson Milo, showcased the AI’s tendency to "correct" reality—in this case, adding a tooth that Milo had recently lost, forcing me to re-prompt the model to reflect his current state.

Official Perspectives and Technical Implications

The technical implications of these projects are significant. When working with Intellectual Property (IP), such as my attempt to create a "Jar Jar Binks" figure, the AI’s safety filters become prominent. OpenAI’s guardrails are designed to protect copyrighted characters, which forced me to pivot toward a generic "space alien" figure.

Furthermore, when I shared my creations with numismatist (coin-collecting) friends, their feedback was immediate: "Those look fake." This serves as a vital reminder that while AI is adept at mimicking the look of an object, it lacks the underlying domain expertise to replicate the intricate details that experts use to verify authenticity.

Implications for the Future of Creative Expression

What does this mean for the future? We are entering an era where anyone can be a toy designer. However, the role of the human operator remains paramount. The AI is the brush, but the human is the artist. As we move forward, the "hallucinations" that currently frustrate users may eventually become a controllable creative variable, allowing for more precise artistic direction.

For those interested in the technical side of these experiments, my ongoing work with figures like Dotti, my VP, and the continued refinement of "The Two-Pack" (featuring Arlene and Dean Torrence) demonstrates that with persistence, the gap between the "first attempt" and the "final version" is closing.

Conclusion: A Work in Progress

As I continue to document my journey with these digital collectibles, I am reminded that the value of this project is not in the perfection of the image, but in the stories behind the people. Whether it is a childhood crush on Susan Cowsill, a tribute to the legendary Jan and Dean, or simply capturing the spirit of a friend, these figures serve as snapshots of a moment in time.

I still have a long list of figures to complete, including more refined versions of existing pieces and new, challenging projects. I invite my readers to stay tuned to the blog. There is much more to come, and I look forward to sharing the next set of results as I continue to push the boundaries of what is possible with AI-assisted design.

For those who wish to review the second installment of my design series or view the full gallery of figures created to date, please visit the links provided in the sidebar. Thank you for following along on this journey from the high seas of the Flower Power Cruise to the cutting edge of the AI digital workspace.