Even though the comparison isn't quite accurate, I generally agree. Even though it's much easier now to build your own software, you still have to invest time and have at least some idea of what you're doing. I'd compare it more to cleaning companies. Anyone can clean on their own, but not everyone wants to spend their time doing it.

The question remains: why, then, do so many people not make any money?

Haiku, Sonnet and Opus… How to decide what model to use when?

If one person can now build almost anything with AI, why is it still so difficult to build something people will pay for?

AI is the best thing that could ever have happened to internet meme culture. 😅

Vibecoding killed one of the best parts of programming. That feeling when hours disappear. When you’re completely locked in. If you miss that: Start writing.

AI video looks cinematic. Most of it says nothing. Founders confuse render quality with sales copy. Slick frames don't move a boring product. Story first, render second. You did the thinking. The model did the polish.

I remember seeing Robert when he had maybe 7k followers and was sitting at Amsterdam airport on his way to Rio, talking about building a $10k/month SaaS. It didn’t feel historic. It was just another post from someone trying to make something work. That’s the strange part. Once someone becomes successful, we start remembering their early stage differently. And maybe that’s also why I like looking back at posts like that. Not because I’m anywhere near where he is now. But because I remember a point where the gap felt much smaller. He was still figuring things out. Still trying to make the business work. Still somewhere in the middle of it. And then a lot changed in a relatively short amount of time. I find that genuinely inspiring. Not in the “anyone can do it” sense. More as a reminder that where someone is today can look almost disconnected from where they were not that long ago.

OpenAI just killed your AI startup. Or did they? DevDay showed once again how fast OpenAI is moving up the stack. Agents. Computer use. Hosted infrastructure. More and more of what startups used to build themselves. Every new release makes some products obsolete. But it also makes entirely new products possible. The founder takeaway: If an OpenAI release can kill your startup overnight, you probably didn’t have a moat. Don’t build around the model. Build around a problem OpenAI doesn’t care to solve.

Do you think OpenAI killed a lot of startups or enabled to create even more profitable businesses?

Everyone's building the same $5 SaaS. Vibe coding killed the barrier. The moat isn't the product - it's real-world data. AI needs ground truth, not another clone. APIs that pipe local business data, customer records, whatever those stay relevant. That's the play.

I want to say something about this because I believe that many people still underestimate this issue too much. In principle, AI has solved a real problem that, for a long time, only company executives with employees had to deal with. For the most part, the average self-employed person hasn’t even realized yet what impact AI has on the way they work. AI can probably handle many repetitive tasks—the kind that absolutely EVERYONE has—better and faster. The problem, however, is that AI can only solve these problems if you, as the user of the AI, already know the task well enough to be able to automate it. And that’s exactly the point. A company executive has learned over time how to assign and teach employees specific tasks so they no longer have to do them themselves. A typical self-employed person, however, often doesn’t know this. The skill lies in being able to break down such a task into clear steps, define the workflow, and evaluate the result. I can only encourage anyone who hasn’t yet truly understood and used AI to learn, develop, and be able to describe these skills. Otherwise, AI will always remain just a tool that seems as if it solves all problems but never actually leads to real results. AI is a tool. You have to learn how to use a tool. Only once you’ve learned how to use it can you use it successfully.

Most tasks never need superintelligence.

It never ceases to be “fascinating” how many of my colleagues can devote an incredible amount of time to getting worked up about their jobs, pay raises, retirement, and vacation, but don’t spend a single second building something that would completely resolve all these issues once and for all.

I just had to pre-order *The Attention Playbook* for $174 because my birthday is on April 17 😄 I can't wait to see what secrets @robj3d3 will reveal. 🤔 theattentionplaybook.com

If building with AI is so easy and fast now, why is it still hard to run multiple businesses at once?

Do you still remember that viral ad that was created entirely with AI? It's already two weeks old.

AI is a product of a very specific point in human history. It could only appear after civilization had spent decades unknowingly assembling all of its prerequisites.