I've spent over $20,000 on API costs per month. Not on subscriptions. Not on wrappers. Just API access to the models that power every tool you're using right now.
I'm not a developer. I can't code. 6 months ago I didn't even know that a terminal existed. But I understand how these tools work better than 99% of people, and that understanding is what got us unlimited deal flow in five months.
(Everything in this article applies outside of filmmaking.)
The Revolution You're Standing In
We are living through a technical revolution. Not an upgrade. A revolution.
The last time this happened was the internet. Out of the internet revolution came Apple, Amazon, PayPal. A few years later: Facebook, Uber, Airbnb, Instagram, Twitter. Companies worth hundreds of billions of dollars that changed how the entire world operates.
Here's what made that moment different from normal progress:
When evolution is linear, new ideas are hard and progress is slow. When there's a fundamental jump in technology, something different happens.
Thousands of ideas that were easy to think of but impossible to build become buildable.
Social media was an obvious idea. Online shopping was obvious. Video calling was obvious. None of them could exist until the infrastructure existed. The second it did, the people who understood the technology first built the companies that now run the world.
We are standing in the exact same moment. AI is the infrastructure shift. The obvious ideas are everywhere. And the window is wide open.
The Only Thing You Need to Figure Out
Here's what most people miss about the companies that won the internet revolution.
Apple didn't invent the microprocessor. Didn't invent the touchscreen. Didn't invent the cellular network. Steve Jobs took existing core technologies and combined them into one product with perfect UX.
The iPhone wasn't a technological breakthrough. It was a UX breakthrough.
The same pattern is repeating right now. Runway, for example, figured out the UX for AI video.
And the best example of all: OpenClaw
OpenClaw wasn't built with OpenClaw. It was built with Claude Code and the core tools that were already available. The creator wasn't some Einstein super-inventor.
He was a man traveling around, building tools for himself. Tools that sped up his own day-to-day workflow. He used the raw APIs, the terminals, the core infrastructure. He kept building until the tools were useful enough to share with other people. Two or three months later, OpenAI bought him.
This is the pattern. And it removes enormous weight from your shoulders. You do not need to invent core technology to win this revolution. You need to figure out the UX. The layer that wraps existing technology into something people actually want to use.
That's the only thing. And the window for figuring it out is open right now.
Why Wrappers Will Cost You Everything
So here's the real point.
OpenClaw wasn't built with OpenClaw.
And whatever you're going to build that actually makes a difference won't be built with a wrapper either.
Every minute you spend inside a polished interface that hides how things actually work is a minute you're not spending understanding the mechanics underneath.
Understanding gives you leverage. And leverage is the only thing that compounds.
I am not a technology guy. I don't read whitepapers. I don't write code. What I understand is how these tools think.
What they prioritize. What they forget. When they break.
That understanding, not the ability to click buttons in someone else's product, is what separates people who use AI from people who build with it.
The Loop
Here's how you actually build that understanding.
Spend a little money on an API. Not a subscription to someone's app. Direct access to the model.
Use the tools to learn how the tools work. Not to be productive. Not to ship something immediately.
Increase the quality of your instructions based on what you learn. Same model, same compute, same cost. Better instructions, wildly different results.
The gap between a generic prompt and a precise one is the entire gap between amateur and professional output. I've watched it happen a thousand times.
Sell what you build. Not the tool. Not the wrapper. The specific automation, the specific workflow, the specific thing you built that saves real time and produces real results.
With the money you make, spend more on the tools. Go deeper. Continue the loop.
This is a compounding cycle. Every rotation, your understanding deepens, your output improves, and your value increases.
How to Actually Learn These Tools
Use the tools constantly and test everything. Don't follow someone else's tutorial. Push the tools in directions they weren't designed for.
Find the limitations. Find where the model breaks. Find the exact moment where context length starts impacting quality, where the model starts compacting information, where it decides what to keep and what to throw away.
The breaking points teach you more than a thousand successful prompts ever will.
Don't try to learn the code. Try to execute. Push until something fails. Then study the failure. When context gets too long, what does the model remember? What does it forget? What does it think is important enough to hold onto when it's running out of space?
When you figure that out, you understand more about how these systems work than 99% of the people using them every day.
Understand memory. These tools have memory systems. Files that persist between conversations.
When those files get too large, the model starts losing things. Learn when that threshold hits. Learn what to remove and what to keep. Learn how to prune your memory without losing what matters.
This is information architecture. It barely existed a year ago. It's becoming one of the most valuable skills in the world and almost nobody is talking about it.
Read the thinking process. When you ask Claude a complex question, the model generates a chain of reasoning before it gives you the answer. Most people skip straight to the response. Don't.
Read the thinking. Copy the entire thinking process. Paste it into Grok. Run it in the new beta mode.
Ask Grok to break down what Claude was actually doing, step by step. What assumptions it made.
Where it was uncertain. What it prioritized. Use one model to reverse-engineer the other.
You will learn more about AI reasoning in one afternoon of doing this than in a month of watching tutorials. This is not a metaphor. This is how I learned what I know, and it's available to anyone.
The Board Meeting
This is the advanced play. The thing that made me realize how deep this goes.
I run board meetings with AI agents. Not metaphorically. I take multiple agents, each loaded with different knowledge, each with a different area of expertise, and put them at a simulated table where they communicate with each other.
I built an agent trained on every post I've ever put on X. Every engagement metric. Every momentum pattern. What content performed, what died, and the reasons behind each.
I brought that agent into a room with my orchestrator, the main system that holds my operational memory and runs my entire studio.
I let them have a ten-minute conversation.
What came out of that conversation was something I wasn't prepared for.
Insights I hadn't seen. Connections across completely separate datasets that I hadn't made. Strategic recommendations sharper than anything I could have produced by myself, thinking for hours.
This is what understanding the tools unlocks. Not slightly better responses. Not faster output. An entirely different relationship with the technology.
You don't get here by using a wrapper. You get here by understanding what's underneath.
The Progression
Start building automations for yourself. Small ones. Something that saves you twenty minutes a day.
Nothing fancy. Just something that works.
Share it with a friend. Everyone wants twenty extra minutes. That's the easiest sell in the world.
Nobody says no to getting time back.
Get better. Turn twenty minutes into an hour. Share with more people.
Show it on X. Start selling your specific, extremely specific automation to the people who need exactly what you built.
The progression is always the same: understand, build for yourself, share, sell, compound. And it always starts with the same decision. Drop the wrapper. Use the core tools.
You never know when you build the one automation that changes everything.
The guy who built OpenClaw didn't know that building tools for himself while traveling around Morocco would end in an acquisition by OpenAI two months later.
He wasn't trying to build a company. He was trying to understand the tools.
The Window
OpenClaw wasn't built with OpenClaw. The iPhone wasn't built with an iPhone.
Every tool that looks effortless on the surface was built by someone who understood the hard stuff underneath.
We are living in a technical revolution. The last one created trillion-dollar companies and a generation of people who saw the infrastructure shift before everyone else and built on top of it while the rest of the world was still figuring out email.
This revolution will create the same thing. And the window won't stay open forever.
The only question is whether you're going to build with the core tools or keep renting someone else's understanding of them.
I have a TG channel where i share thoughts like this. for the rest, i don't really have anything to sell you besides the films that we make.
euhm. oh yea, this is a transcribed voice message turned into an article, i don't like to write.