Shouting at Claude to build my ideas
August 24, 2026 · 10 min read
I hated programming. That is not really true, so let me correct it before I go any further. I did not hate it, it just never felt like play to me, and that turned out to matter far more than whether I was any good at it.
I built a software company anyway. I took Nalikes from zero to a million dollars in annual revenue over five years, and in all that time I never wrote a single line of code. I did not even have access to our clients' GitHub, because that was not my job. There were three of us, and while my two partners built the products and delivered the work, I was the one responsible for the revenue.
So I have spent five years shipping software without ever building any of it with my own hands. Last year that started to change, and a few months ago it changed completely. Today I build my own ideas by talking to Claude until they exist, and the first one I built by myself has already been sold.
To explain how I got here, I have to go back to when I finished college.
I came out with good grades and absolutely no idea what to do with them, so I called one of my uncles, who was the most successful man in our family, and asked him what I should study. He told me to look into something with computers. When I sat down to fill in the online university application, I chose programming.
That was the entire process. One phone call with one uncle, and I am still grateful to him for putting me on that path, because nothing else in my life at that point was pointing anywhere.
You have to understand how little any of us knew. I studied under the federal board, where the textbooks had not been updated in decades, so in eleventh and twelfth grade we were still being taught what a mouse is and what a keyboard is while the real world moved at a speed none of us could see. University was not much better. Nobody there was telling us what was happening in technology or in Silicon Valley, and everything I eventually learned about the industry I picked up years later on my own, through crypto. That gap between what schools teach and what the world is actually doing is a separate problem, and I will write about it another day.
I figured out fairly quickly that I could learn programming, but I was never going to sit and write code all day, because that was not the future I imagined for myself.
So I stopped. There were other reasons, though that was the main one. I wanted to make money, so I went and learned the skills that made money faster .
Around that time the three of us started entering hackathons, and it became obvious that somebody had to pitch and somebody had to sell. I was already running a small ghostwriting agency on the side, which had taught me the basics of business, so the decision made itself. My two partners would go deeper into programming, and I would go deeper into sales.
I am glad I did, because Naval says learn to sell and learn to build, and if you can do both you will be unstoppable. I think I have ended up with both, at least to some extent.
For the next five years I did everything a business needs except build it. I signed clients, grew the revenue, hired people, and made sure there was always work coming in, while my partners handled everything technical.
That does not mean I knew nothing. I understood how software gets built end to end, and I was technical enough to sit with a customer and solve their problem. My real advantage was that I could speak both languages. On a call my partner would go deep into technical detail and a non-technical founder would quietly lose the thread, and I was the one who understood both sides of that conversation. IQ on one side, EQ on the other, and looking back that is probably the single biggest reason the business grew the way it did.
Then last year the vibe coding tools started showing up, Lovable , Replit , Bolt and many others. I tried them and found that they were good, though nowhere near good enough to replace a real developer or a development company. What I could not shake was the feeling that programming was about to change forever and that this would eventually disrupt our business, so we told our developers to start using Cursor and paid for their memberships. I wrote about it at the time.
Vibe coded this simple game using @lovable in <10 mins. Try it now - crypto-snake-quest.vercel.app Been building apps using @lovable, @boltdotnew, @v0 and @cursor_ai for the past few weeks. Amazed by how great these tools are. These tools have made us reevaluate our Show more
After that I put them down. My priorities were somewhere else, and as an individual I did not really touch them again for months.
A few months ago my timeline started filling up with the same conversation from every direction. MCP servers, new models, new tools, and then around February the new Opus model arrived alongside everything else, and people were not talking about it the way they talk about a shiny thing. So I sat down and experimented again.
The difference between one year earlier and that afternoon was not small. These tools were a hundred times better.
After that I got obsessed. I used them for building, for research, for writing and for almost everything else, and I ended up spending more of my day talking to Claude than to any human being. I even tweeted about it at the time. Most of my waking hours were going into the Claude mobile app, the desktop app and the web app, and somewhere in the middle of all that I moved off ChatGPT completely, because the difference was obvious.
I’ve talked to more AI agents in the past few weeks than humans
Most people still use these tools like a search engine. They type a question, read the answer and close the tab, and I was doing exactly that last year, so this is not a criticism of anyone. What I understand now is that these are the smartest things any of us have access to, and they should be treated that way. I learned to prompt properly, and then I stopped typing and started speaking, which changed everything, because I can talk for twenty minutes straight and it will take all of it and actually use it.
Every single day something surprised me. I handed over a few thousand of my personal notes and had a proper summary of all of them within minutes. These tools are extremely good at reading enormous amounts of text and telling you what is actually in there, and they are even better at research, because they will pull together information from places I would never have found on my own.
All of this was happening while I was finalising my exit from the studio, and it is the reason I announced I would start making content , because instead of a big team behind me I now had something that could help me think, plan and build on its own.
Somewhere in the middle of that, I started shipping.
I tested a lot of ideas. Browser extensions, small tools, half-finished apps. Building products for five years teaches you one thing very clearly, which is that the idea is almost never the point, and what matters is whether anybody reacts when you put it in front of them. So the plan was simple. Ship the basic version and see what happens.
One random day I made a tool that finds the Telegram handles of crypto founders, and I posted about it . The post did around two thousand impressions and a lot of people asked to try it. What I had actually posted was a front end prototype with no backend behind it at all, recorded with a browser extension called Cursorful , and it went much further than I expected.
We scaled @NalikesStudio to $1M ARR through cold outreach. We've done cold outreach on X, LinkedIn, email, Discord, and Telegram. But over the past few months, Telegram has been getting us the best results by far. The problem? Finding Telegram usernames of decision makers. Show more
So I spent a few more hours making it real. I improved the tool, started sharing it with people, watched it get used, and added a payment gateway. It had limitations and it was a long way from perfect, and people still wanted it.
Then I had some health issues and stopped working on it altogether.
While it was sitting there, somebody from the industry reached out. They wanted the tool, the data inside it and the technology behind it, and they bought it from me .
SaaS Acquired!!! Another tool to help crypto BD reps win (below) I'm betting that we'll see more sales people in the crypto industry in the future. I want to give them everything to succeed so they stay and their projects stay alive! Sales > Everything
I'm acquiring my first SaaS! Have AI find a crypto Telegram handle in 10 seconds instead of manually searching Why? Data is THE moat in this AI world - I'll grow the database - Plug it into TGcrm If you're a BD rep, comment and I'll give you FREE credits
I did not make a fortune out of it, and that was never the point. I made enough to prove to myself that this is worth taking seriously.
That is where this stopped being interesting and started being obvious.
Think about what used to happen when I had an idea. I would brainstorm it with my cofounders, then we would write the requirements, then wireframe it, then hand it to a developer or build it ourselves, and weeks or usually months would pass before there was anything to look at. Today I have an idea in the morning and an MVP by the afternoon, and I can put it in front of real people the same week and find out whether anybody wants it. Anyone can do this, and if you can speak English you can do this.
Which brings me to the title.
There is vibe coding, and there are ten other names for the same thing, though I think of what I do as shouting at Claude.
I talk a lot, and anyone who knows me will confirm it. I send voice notes that visibly upset people, and I tell them the same thing every time, which is that this is how I communicate best. It turns out the models prefer it too, because they want the whole burst of information with the mess left in.
So that is the method. You open a session, say what you want to build, and you keep talking. It gives you a version, and then you shout at it. Fix this. No, not like that. Do it again. Sometimes you tell it that it is being incredibly stupid, and then you carry on.
That is the entire technique, and I might try to coin the term.
Here is what I am actually doing with it. For years I had ideas I never built, because my developers were busy with client work and I had sales numbers to hit, and that excuse is gone now. So I am running an experiment where I build the ideas myself, and every one of them goes on my ideas page .
The rules are simple. If an idea excites me I build it, if it gets traction I keep going, and if it does not I scrap it and move on. A domain costs a few dollars, a plan costs twenty dollars a month, and posting on Twitter is free.
The first one sold, which might be beginner's luck. If I can do it a few more times then it becomes a process, and then I can show other people how to do it too.
I do not know exactly where this goes, though two things look obvious to me.
Good developers are going to produce far more, at higher quality, in less time. People who were always intimidated by programming are going to build things for the first time, and I say that because for five years a lot of our clients were non-technical founders who were terrified of building anything themselves. Those people now have Lovable , Base44 , Claude , Codex and a dozen others, and they are going to build.
I could be wrong about all of it, so do not take my word for it, go and find out yourself.
One last thing, because it comes up every single time I talk about this.
Technical people are the most sceptical, and almost every developer I speak to tells me these tools are fine for prototypes and nowhere near production. I understand it, though I also think a lot of it is fear, because they spent years learning a skill and something has turned up that does a chunk of it. The ones thinking further ahead are not arguing about it at all, they are using them and shipping more than they used to.
When the calculator arrived people said the same things, and when the printer arrived the scribes said the same things, and they were right that a job was disappearing. What they missed is that the job was never the point. The arithmetic was not the work, the copying was not the work, the thinking was the work, and it still is.
So this is the best time there has ever been to have an idea. For twenty dollars a month you have a very good developer sitting there, waiting for you to explain what you want.
Go and shout at it.