There's a reason 90% of startups blow through their marketing budget and get nowhere.

Helping brands go public and also raise tens of millions of dollars and do hundreds of millions in revenue, one of the questions I get asked the most is how to spend (marketing) budget at each phase of company growth.

What I find is that typical companies want to build additions to their house before they’ve built a house. They want to build a skylight before they’ve poured concrete in a foundation. It’s easy to get ahead of ourselves…

It doesn’t make sense to throw ad spend on a company with no brand and no consistent messaging.

It doesn’t make sense to start with SEO in a deep tech environment, where the world doesn’t even know what it is you’re offering yet. People don’t search for new technology the way they search for “discount dentists near me”.

The way I’ve helped so many start-ups and scale-ups get to the next level is by ruthlessly focusing on sequencing. Yes, some day, you will want to do everything. But today? You need to get your branding, your messaging, and your assets right. Without those things, you can blow through your budget and get nowhere.

There’s a reason unicorn start-ups all tend to work in a certain way. And that way is almost always: brand first.

Your brand is either a liability or an asset that will pay dividends for years to come.

The order matters.

A friend told me coding now feels like shaping clay on a pottery wheel. I haven't been able to stop thinking about what that means.

Computers: Cold, sterile, lifeless, inside-the-box, boring.

Real life: Flowing, messy, organic, unpredictable.

In computer land of the past, programs were traditionally made by thinking through problems in a highly mathematical, organized, and rigid way.

The way living organisms evolve is very different. For example: Why do humans have sinuses that drain from the top of our nasal cavity instead of the bottom? It only makes sense if we see how we evolved from creatures who didn’t walk upright. How many of our terrible colds would have been alleviated if we could just design ourselves the way we engineer software?

But now that’s changed.

Gone are the days of thinking through an app from start to finish and coding it according to a plan. One friend of mine said coding now is more like shaping clay on a pottery wheel. The code is ever-evolving in front of our eyes, and it’s our job to shape the clay as it spins.

As we all build the software tools that will define the next 10 years of our personal lives and careers, we’ll be continually molding the software programs that govern our lives. We’ll watch them grow, improve, and change just like a living organism.

We don’t know where the organisms we make today will end up in a couple of years, but we know that, like living objects, they will be flowing, messy, organic, and unpredictable.

For the rigid thinkers, this is bad news. But for those who’ve wanted a taste of the power programmers have always had without becoming inside-the-box and boring, it’s an incredible time.

If AI can do everything all at once, why have I had to code with it 7 days a week for the last 6 months?

What I’ve learned is: Even with AI, it still takes an enormous amount of time to work through complex problems.

Yes, not writing the code saves us time on a certain kind of problem, but we are a long way from AI doing everything for us—even in software.

As our (coding) projects grow, the number of decisions we need to make also grows. It’s exponential.

What makes your app yours are the thousands of tiny decisions you’ve made daily over a period of months.

There’s no foreseeable reality where agentic AI will just go off and build what you want without you guiding it in a meaningful way.

Unless what you want is so basic, so generic, and so commodified that there was no point in you building it in the first place.

It's to train our brain.

If you're overwhelmed by AI this-that-and-the-other right now, you might be missing the point: Right now, we get to improve our brain’s neuroplasticity like never before.

Here's the thought experiment that changed how I see this:

"What would you do with a million dollars?"

It’s a deceptively hard question right? Most people can’t immediately answer it. And before you answer, remember to quote Office Space: “You don’t need a million dollars to do nothin man. Look at my cousin, he’s broke don’t do s***!”

Now try: "What would you build if you could build anything?"

Equally tough.

That's the real challenge of this moment. It’s learning the tools that’s hard—it’s learning to think bigger because the tools now let us.

Yes, there's a way to use AI that shrinks your brain:

• Cheating on exams as a student
• Cheating on professional work as an adult
• And cheating on your spouse with LucyGPT.

That dumbification is real. I won't pretend it isn't.

But here's what I've noticed building with AI daily:

Solving complex problems—architecting systems, connecting patterns across domains, debugging logic I couldn't hold in my head alone—activates my frontal cortex in ways that feel like growth, not atrophy. Like solving a wonderful crossword.

Some skills I used to have? Probably fading. I don't bust out my abacus much these days.

But we all get to see the bigger picture like never before. Our brains can architect entire systems instead of paragraphs.

The trade-off isn't "smart vs. dumb."

It's "which kind of thinking are we training?"

In the future, we’ll all have highly personal apps that do our very specific tasks, professionally and personally.

These won’t be off-the-shelf tools like Gemini, Claude, or ChatGPT.

They will be tools that we build using these off-the-shelf subscriptions.

Why is it important to start automating tasks that are repetitive and wasting your time right now, knowing that these models will only improve?

Let me use an analogy:

If you are a lucky homeowner, you know the difference between the bones of your house and a remodel job.

The bones are what you buy. The ability to remodel is the ability give your home a facelift every time $50,000 is burning a hole in your pocket.

Given unlimited resource, I’m sure that every homeowner wouldn’t mind a regular remodel, an easy way to keep their home fresh for decades.

Well, that’s how AI works.

You use the tools available today to build the bones of your app. You set up the databases, the structures, and the concepts behind the personal apps that automate your life and business. You accept that these apps are imperfect works in progress.

Here’s the beautiful part: each time a new model drops, with greater capabilities, you get to have it go through your code base and improve, fix errors, and apply a new coat of paint. In this way, you can improve your house of code every time a new model drops, often in one fell swoop.

I’ve built over 20 apps at this point, and as soon as Claude announced Opus 4.7, the first thing I did was have it run through my codebase, update, refactor, and polish. This is how my apps are constantly improving and becoming more functional and useful over time.

But you can’t remodel a house that doesn’t exist. You can’t improve on an app that you haven’t built.

So today is the day to build the foundation of your house. And trust that you will be able to remodel it perpetually over the years. Just think of what kind of palace you’ll have five years from now, if you start today?