"Chaos Detective," a Personal Experiment Between AI, Imagination, and the Beauty of Uncertainty
How my reflections on AI ethics led to my first app, a playful complement to the Custos AI Framework
Have you ever had an idea so strange that it made you think 'either this is completely wrong or it might work'? That was precisely my feeling when I considered transforming a rigorous methodology, conceived to pre-empt serious algorithmic disasters like entrenched societal biases or runaway misinformation, into a game where the explicit goal is to creatively lose oneself in the ensuing chaos. It could have been dismissed as madness, yet it was this very juxtaposition that I felt compelled to explore.
It all started with the Custos AI Framework, a conceptual toolkit I’ve been developing and sharing on 'Learn Vedanta Substack.' At the heart of this framework lies the concept of Preventive Chaotic Intelligence that I theorised: a methodology envisioned to operate by generating and analysing a vast landscape of potential futures, hypothetically exploring tens of thousands, perhaps even up to a hundred thousand, distinct variations of an algorithmic scenario. Such a deep dive would aim to meticulously map how even subtle, almost imperceptible differences in initial conditions or input data could ripple outwards, leading to wildly divergent, and sometimes ethically problematic, outcomes. The core idea behind this theoretical approach is to identify and thereby help neutralize 'negative attractors' – those specific zones within the imagined landscape of future possibilities where algorithmic decisions are most likely to degenerate into discrimination, amplify misinformation, or trigger other significant ethical violations – ideally before these issues can manifest and cause harm in the real world.
But while immersed in developing these critical prevention mechanisms, a tantalizing question began to tug at my mind, a question that simply wouldn't leave me alone: what would happen if I applied the spirit of this logic—navigating complex, emergent patterns derived from myriad possibilities—not for defense, but for pure, unadulterated creativity? What if, instead of diligently searching for theoretical negative attractors to pre-emptively shut down, I repurposed the approach to discover and explore vibrant 'creative attractors' instead? I became captivated by the challenge of finding a light-hearted, engaging way to offer a tangible taste of this underlying philosophy through a deceptively simple game mechanic."
"Chaos Detective": When Theory Meets Game
This is how "Chaos Detective" was born, a sort of playful complement to the Custos AI Framework. Not a game in the traditional sense, but an exploration space where uncertainty becomes a source of creativity instead of anxiety.
The idea was to create a "chaos detective" who moves in a world where information is fragmentary and truths elusive. The player faces a deliberately open investigative case, where they must use imagination as their main tool.
The mechanic was crystal clear:
An initial problematic scenario
Five turns in which the player adds reflections, questions, and intuitions
At each human input, the AI generates ten "chaotic echoes" - not answers, but creative variations that deviate from the course of the investigation
At the end, no revealed truth. Only "Chaotic Epilogues" and an analysis of the emerged complexity
It was Preventive Chaotic Intelligence - the concept I theorised - translated into playful form: from chaos control to dancing with uncertainty. A bit like transforming an anti-theft alarm system into a disco.
I had the vision, I had the mechanic, I even had the cyberpunk aesthetic in mind - those dark tones punctuated by luminous accents that evoke a digital world where meaning hides between the lines of code.
What I didn't have was the slightest idea of how to translate all this into a functioning application. I had never written a line of code in my life. It was like having the perfect project to build a Ferrari and discovering you don't even know how to change a tire.
The Team: Two AIs, Two Roles
The breakthrough came from dividing the work between two artificial intelligences, each specialised in its domain.
Qwen3-32B-A3B (an open-source AI model) fit perfectly into the game mechanism as a generator of "chaotic echoes." Its task is to transform each player's input into ten unpredictable creative variations. The beauty of the system is that the better the prompt is written, the more surprising the echoes are and capable of opening unexpected narrative paths. It became my accomplice in the art of controlled unpredictability.
Gemini 2.5 Pro acted as an impeccable technical translator, transforming my poetic descriptions - "I want a dark box with flickering text to appear here" - into HTML, CSS, and JavaScript that actually works. A true saint of digital patience.
My role was to act as a bridge between the original idea and its realisation. A bit like a translator working between different languages: concept, game philosophy, content, and then coordination between the two AIs to make everything fit together in the right way. Not the most glamorous job in the world, but fundamental for transforming an abstract theory into something people can touch and experience.
The Experiment in Action
"Chaos Detective" became the laboratory where the principles of Preventive Chaotic Intelligence transform from governance methodology into a tool for personal exploration. The player experiences on a small scale what it means to navigate complexity, where every input generates unforeseen ramifications, where the goal is not to find the answer but to explore the landscape of possibilities.
There was something deliciously circular about all this: I was using artificial intelligence to create a tool that invites people to reflect on their interaction with artificial intelligence. A bit like using a mirror to look at how we look at ourselves in the mirror.
The Result
What emerged from this triangular collaboration is a small experiment that celebrates uncertainty instead of fearing it. Not a game to win, but a space to inhabit. An invitation to formulate bold hypotheses, to ask uncomfortable questions, to step off the beaten paths of conventional thinking.
It's a little game that needs improvement, but I hope it conveys at least a taste of what I mean when I talk about intelligence that doesn't seek to control chaos, but to dance with it together to discover where it can lead.
If you're curious about the idea of playing with uncertainty, exploring how human imagination dialogues with artificial intelligence, and experiencing a small practical application of Custos AI principles, you can try "Chaos Detective" here:
https://chaos-detective-llm-enhanced-edition-475351370946.us-west1.run.app/
Every piece of feedback will be valuable in understanding whether this approach can open new paths in AI-assisted creativity. And if the game seems strange to you, well... that was exactly what I hoped for.
"The question is in the answer; the answer is in the question."
Source: Shibayama Zenkei, in his commentary on the Mumonkan (The Gateless Gate), a foundational collection of Zen kōans.
With gratitude,
Cristiano
LinkedIn: https://www.linkedin.com/in/cristiano-luchini-1026aa17/