skip to content
Navid Mesbah
فا
← Back

Creative Intelligence Belongs to Everyone

/ 10 min read

Three Sides of One Table

Something changed at Charsook this year that I didn’t plan for. Our clients started arriving with AI already in their hands.

A client used to come with a brief: a few paragraphs, maybe a mood board, maybe a sketch on paper. Now they often come with a prototype they made with AI the night before. A rough interface, a draft video, a generated set of characters, a list of ideas an assistant helped them organize. It’s rarely finished, and it’s often not quite right. But it’s real, and it’s theirs.

On our side, we use AI everywhere: agents write much of our code, help us explore designs, draft tests, and prepare variations faster than any team could by hand. And the product itself, Charsook, is a place where people and AI make creative work together.

So the work now happens in a triangle. The client, Charsook, and AI. All three sides use AI, and AI also sits at the table as a participant in its own right.

This changes the conversation in a way I find beautiful. A client no longer has to describe what they imagine and hope we understand. They can show it. We no longer have to guess for weeks before showing something back. We can respond in an afternoon with three versions, each one closer. The distance between an idea in someone’s head and a thing on a screen has collapsed.

What Our Role Becomes

When clients can make drafts themselves, people sometimes ask what we are for.

The answer, I think, is that our value moved from production to judgment. A client’s AI-made prototype is a beginning, full of energy and intention, but also full of accidental decisions: structures that won’t scale, interactions that look good once and confuse on the second use, motion that feels nice in a demo and tiring in daily life. Our job is to see the real idea inside the draft, protect what is alive in it, and build the foundation it needs to become a product.

In a sense, we became translators between intention and craft. The client brings intention, more clearly than ever because they could already try it. AI brings speed and endless variations. Charsook brings the judgment of people who have spent more than fifteen years building things, making mistakes, and learning what lasts.

None of the three is enough alone. A client with AI can make a prototype but often can’t make it durable. AI alone has no idea what matters. A studio without AI is too slow for the way people now think. Together, the triangle works.

Meta-Programming: Working One Level Up

The second change is quieter, but for me as an engineer it may be the most important one.

For years I’ve believed in functional programming. Pure functions that always give the same output for the same input. Immutable data that can’t be changed behind your back. Types that describe exactly what can and cannot exist, so impossible states can’t even be written. Small functions that compose into larger ones like sentences built from words. Explicit handling of errors and side effects instead of hoping nothing goes wrong. Property-based tests that check rules across thousands of inputs instead of a few hand-picked examples.

These practices make software calmer. Bugs become rarer and easier to find. Code becomes easier to change without fear. I’ve seen this in OCaml, in Rust, and in carefully typed TypeScript.

But honestly, for most of my career, it was hard to make these practices real in a team.

It wasn’t because the developers I worked with weren’t smart. They were. But functional programming asks for discipline and a different way of thinking, and under deadline pressure that discipline is the first thing to go. Modeling every state with precise types takes time. Writing property-based tests takes practice. Keeping data immutable sometimes means more boilerplate. One developer would follow the principles carefully, another would take shortcuts, and the codebase would end up as a patchwork. I spent a lot of energy reviewing, explaining, and gently insisting, and it never fully stuck.

AI changed this, and the reason is meta-programming.

Meta-programming means working one level above the code: writing the rules, patterns, and specifications that produce code, instead of writing every line by hand. With agents, this became our everyday reality. We describe our principles once, clearly: how we model data, how errors flow, which effects are allowed where, how every module should be tested. Then agents apply those principles everywhere, consistently, without getting tired and without taking shortcuts at six in the evening.

The boilerplate that made functional style expensive costs almost nothing now. The precise types that used to take an afternoon take minutes. The property-based tests that people skipped are written every time. The patchwork is gone. For the first time, the codebase looks like the principles we always believed in.

This is something individual developers on my teams struggled to do for years, through no fault of their own. Now the whole team works at the level of principles, and the principles are honored in every file.

Meta-Learning: Learning How We Learn

Next to meta-programming there is meta-learning: improving not just what we know, but how we learn.

Every time something goes wrong at Charsook, we ask two questions. What was the mistake? And what rule, test, type, or eval would have caught it? The answer to the second question becomes part of the system. It goes into the specifications agents follow, into the eval sets that check quality, into the types that make the mistake impossible next time.

So the team learns, and the system learns from the team. A lesson learned once is applied forever, by every agent, in every new piece of work. That is a kind of organizational memory that was very hard to build before. Knowledge used to live in the heads of a few experienced people, and it left when they left or got tired. Now it lives in the structure of how we work.

Meta-learning also changes how individual people grow. When a developer sees an agent apply a functional pattern correctly, with an explanation, across many examples, they learn the pattern faster than from any lecture I could give. When they disagree with the agent and turn out to be right, they learn something about their own judgment. The tools became teachers, and the developers became better architects.

Free, Not Busy

Here is the danger I see, and I’ve felt it myself.

When AI makes everything faster, the easy response is to do more of everything. More features, more meetings, more messages, more output. Fill every saved hour with new tasks. Many teams and many people are doing exactly this, and they’re exhausted. AI made them busier, not freer.

I think this is the wrong use of the gift.

The real gift of AI is time to think deeply. Hours that used to go to boilerplate, repetitive fixes, formatting, searching, and mechanical work can go to the questions that actually matter. What should this product really be? What does this client actually need, beneath what they asked for? What is the most elegant structure for this problem? What would make this tool a joy to use instead of merely useful?

These questions need slow, uninterrupted attention. For most of my career I rarely had enough of it, because the urgent mechanical work always pushed it aside. Now I try hard to protect it. When agents take the mechanical work, I don’t fill the gap with more tasks. I take a walk, sketch, read, play music, or sit with a problem until it opens.

The most important work I’ve done this year came from those quiet hours, not from the busy ones.

The Essence of Life Is Creative Intelligence

This brings me to the idea at the heart of Charsook.

I recently read a short piece by Susan B. Eirich of the Earthfire Institute, written from a conversation about nature, conservation, and creativity. It opens with a sentence I keep returning to: “The essence of Life is creative intelligence.” She describes creativity as the source of resilience, the ability to imagine a new path when the old one is blocked. And she makes a point I find very important: the power to create is in everyone, not only in people we call artists, and each of us has our own creativity to offer.

That is exactly what we mean when we call Charsook a creative intelligence platform.

For most of history, creative work was reserved for a small group of people with rare training, rare tools, or rare free time. Painters, composers, designers, filmmakers, engineers at the frontier. Everyone else was expected to follow instructions, fill roles, and repeat. Not because they lacked creativity, but because life gave them no room for it. Their days were full of necessary, mechanical work.

AI can change this, if we use it well.

A shopkeeper can design their own brand and tell their own story. A teacher can build the interactive lesson they always imagined but never had time or skills to make. A farmer can model their land and try different plans before planting. A student can turn a question into an experiment instead of a memorized answer. An engineer can explore ten designs instead of defending one. A grandmother can turn her memories into a film for her grandchildren.

None of these people are “creatives” in the professional sense. All of them are creative, because life is creative, and they are alive.

Why This Matters for Society

I believe many of the problems our societies face are, at root, problems of blocked creativity. People who feel they have no way to shape their lives become afraid, passive, or angry. People who can imagine and make new paths become resilient. The same article notes that fear narrows the mind while creativity lights up all of it, and I think that holds for communities as much as for individuals.

If AI is used only to make a few people more productive and everyone else more replaceable, it will deepen that fear. If it’s used to free people from mechanical work and give them room to create, it can do the opposite. It can turn more people from consumers into makers, from followers of instructions into authors of their own solutions.

That is the future we’re building toward at Charsook. Not AI that creates instead of people, but AI that gives people back the time, tools, and confidence to create for themselves. The triangle of client, AI, and Charsook is a small version of this: everyone at the table brings their own intelligence, and the result belongs to all of them.

What I Hope For

I hope we use this moment well.

I hope engineers use AI to finally build software the way they always knew it should be built: principled, well-typed, carefully tested, calm. I hope teams use it to learn faster and remember longer. I hope everyone uses the time it gives back to think, not just to do more.

And most of all, I hope creativity stops being a profession and becomes again what it always was: the ordinary, essential way living beings meet the world. If AI helps us remember that, it will have been one of the best things we ever made.