Prompt Engineering as Ars Alchemica

The manipulation of words in search of new meanings, through magical formulas we know as prompts.

"Nothing is created, nothing is destroyed: everything is transmuted through the prompt."
— Cyber-hermetic fragment, dated 2025.

This digital scroll is for those who are not guided by their insecurities and who do not fear using technology to enhance their reasoning, writing, or productivity. There is no shame, no sin.

Artificial intelligence will not replace human creativity and intellect, but it will render them irrelevant—by changing their nature. The future belongs to those unafraid to become proto-cyborgs, integrating their humanity with technology, sculpting their destiny in silicon.

Here, we’ll explore how to tame our AI assistants to uncover new ideas, novel linguistic scaffolds, or let ourselves be carried away by refined symbolic hallucinations.

The Magnum Opus of this cyber-alchemical lab isn’t about working with matter in alembics and sulfurous compounds to seek the Philosopher’s Stone, but about manipulating words in search of new meaning, through magical formulas we call prompts.

This is my personal contribution on the topic of prompt engineering, derived from a year’s worth of experience with ChatGPT and reading various field manuals.

On Prompt Engineering

In this article, we’ll explore best practices for building prompts and the main techniques of prompt engineering:

• Zero-shot & few-shot prompting

• Role & contextual prompting

• Chain-of-thought prompting

• Self-criticism prompting

Before diving into the technical part, I’d like to discuss how I personally use AI (ChatGPT) for my own brainstorming process.

Let’s be clear. I don’t like the term prompt engineering, because it evokes mechanical rituals and dry technicalities. For me, dialoguing with models like ChatGPT is a form of philosophy. What is philosophy, after all, if not the art of crafting questions and answers?

A prompt always begins with a question we ask ourselves. But arriving at a meaningful question isn’t always easy. This too is a form of problem-solving: sometimes, in the very process of crafting a precise and detailed prompt, you may already find the answer embedded within it. In every prompt (question) lies its answer, waiting to crystallize.

But the act of untangling our tangled thoughts in search of the right question can also happen through prompting itself. A kind of meta-prompting. And so we can use AI to find both the question and the answer, in a kind of cybernetic deconstruction process where both the user and the assistant are in search of meaning.

This view reflects the vibe coding approach—mocked by any self-respecting developer—as it’s more a nerdy artform than software development. But I find it useful as a method for discovering and transmuting ideas.

The Transmutation of Ideas

I’m not much of a thinker. I don’t spend hours pondering big philosophical themes—I’m too lazy and not cultured enough. What I usually write about comes to me as brief intuitions at random moments. For the past year, I’ve used ChatGPT to dig into those intuitions and uncover the questions and answers hidden within them.