‍

I was inspired to build a fractal explorer with AI that also reacts to music (of course), after reading Deep Simplicity by John Gribbin.

This book is an excellent primer on chaos theory and how complex systems can emerge from simple mathematical formulas, including how life itself can emerge from primordial soup!

Fractals are a great example of the kind of structures that are found everywhere in the universe, including possibly the most famous visual representation, the Mandelbrot fractal.

Even though this book was written before AI, it explains exactly the way AI will achieve Recursive Self-Improvement in the near future as a series of logical steps already set in motion.

As a professional designer, I’m much less interested in what generative AI does to replace human talents (what the internet lovingly refers to as AI slop). I’m much more interested in the intersection between mathematics and design, something AI coding tools are uniquely poised to help us with.

How can we generate design concepts at the atomic level using mathematics? This book and a few weekends of messing around with AI code will get us to some amazing new heights here.

The future is about how talented people can think with these tools and have fun with them, not about how these tools can replace what makes us human.

Keep exploring, y’all!

Or else its viewpoint will confuse our own.

When you ask an LLM to write for you, it will feel great until you realize that it’s not actually your voice or what you’re trying to say.

When you ask an agent to design for you, it will look great, until you realize that the design is non-specific and the same as everyone else out there.

When you ask AI to do anything for you, accepting its output as-is is a dangerous trap we all have to fight hard to avoid.

I say we have to fight hard, because our agents are constantly lulling us into sleep. Into accepting their opinions. Claude laid out my webpage like this, so it must be right!

Push back. Prompt philosophically. Use it as a tool, but don’t relinquish command.

Add context. Add nuance to your prompts. Add the “why”.

The danger is to let an agent do the thinking for us, and to let its taste override our own.

Use it as a tool in your arsenal of many tools.

But constantly ask yourself: Am I using AI, or is it using me?

It’s been doing the rounds that someone is giving away clones of all Adobe software, all coded with AI.

And I get it: How many of us yearn to be free from the shackles of every monthly subscription, bleeding us dry, like so many paper cuts?

And to be clear, what Brandon Thomas has done is EXACTLY what smart people should be doing right now, and I applaud him for his ambitious effort, really!

But here’s what people don’t understand yet about building software with AI:

The surface level is just the beginning.

What makes Adobe Adobe isn’t the GUI or the obvious features we can see in Photoshop.

It’s the millions of decisions that have gone into decades of development.

It’s the feedback from millions of designers relying on the tools professionally for their entire careers.

Our regular brains can look at today’s Photoshop and see how its lineage traces back to the very first version, released on Apple computers in 1990.

Software is an organic process. No one could have predicted in 1990 what graphic designers would need in 2027. And yet, many of the design cues from that era are still alive today, just with a different coat of liquid glass on top of them.

Software is a conversation.

And great software is a conversation that’s been running for decades.

So our normal brains understand that commercial software is a process. Our AI mania brains fool us into thinking we can rebuild Adobe’s suite in a weekend.

That’s not true.

We'll never replicate the 30+ years of user feedback, bug reports, and workflow refinements that make the original software truly powerful, or at least, it will take us 30+ years to do it and we’ll have to relearn everything Adobe already knows.

But what IS true is that AI allows more of us to engage in the process of enterprise software building. It democratizes the building of these tools.

Unless you’re willing to devote the next 10+ years of your life to building your weekend clone, one painstaking feature at a time, Adobe will always be worth the money.

AI isn’t a magic bullet. It just democratizes coding, bringing the time and cost down exponentially. It doesn’t mean time, effort, and feedback loops no longer have meaning.

And it takes people like Brandon pushing the envelope of open source software to inspire us all to push harder and dare to dream bigger, and to challenge stagnant corporations.

But we’re all going to learn that the true value of software isn’t the coding. It’s the conversation. And that takes time.

Bad designer: The work is ugly AND doesn’t function well (most live here).

Good designer: The work is beautiful, functions well, but isn’t justifiable with powerful metaphors (most people using AI live here).

Great designer: All the small and large elements form a cohesive whole, in which each element helps tell the larger story (the holy grail!).

The simple question to ask of each component: Why is this element the way it is? Is it random? Is it just a shiny object? Or is there a defensible reason for why it works like that? Why it looks like that?

Do the macro and micro elements work together or against each other? Or was something done just because “it looks cool”?

Apply this ruthlessly, and watch your storytelling get better and better.

I’ve had an agency growing start-ups and scale-ups from zero to millions in fundraising and revenue for ten years, and I was a freelancer for the 15 before that. I’ve worked in and around start-ups for my entire career.

The scariest shift I’ve seen in the start-up landscape? The business in which everyone uses AI to do all the thinking and talking for them.

How can you tell if you’re in this kind of business right now? Easy. All the emails around you just keep getting longer and longer, with no end in sight, as one AI talks to another. What

“That’s a sharp observation, Tina. Here are 14 in-depth ways your email misses the mark.”

I increasingly believe there’s a right way and a wrong way to use AI. The right way is coding and automating mundane, non-human processes.

The wrong way is claiming that you wrote something that AI wrote for you. How many LinkedIn newsletters haven’t written a word in months? How many conferences have devolved into a series of AI-made presentations that all look the same?

A business today in which every employee is just one ChatGPT talking to another ChatGPT may feel productive, because all the conversations look so serious.

But GPT-speak is just papering over real problems much like a shiny new tattoo won’t cover up my beer gut. Sorry bud, you’re here to stay.

You can’t serious-speak your way to a successful company. You have to build something of real value.

And marketing a real business is about more than just shoving a bunch of semi-technical jargon onto an AI-made template. It’s about communicating an idea effectively and here’s what people tend to forget: communicating it simply.

Sadly, we’re all going to have to send our BS detectors into overdrive over the next several years, as more and more people around us get suckered into the dark side of AI.

I’m just glad I’m not a VC right now.