I had a pretty big outsource software development company, and on top of that I also had technical background, not only market and management knowledge. Before this project, I already had one startup that successfully failed after I spent more than $200,000 on it.
A big part of that failure was team-related.
When I started building goundetected, I was again looking for people. I needed engineers with enough experience, especially in desktop development, and I just couldn’t find the right ones. So instead of wasting more time searching, I decided to try agentic coding and build it myself.
Models were not as good back then as they are now. So I had to go deep into the code almost like I did 10 years ago. I was asking a lot of extra questions all the time: why this solution, why this component, why do we store it this way in the database, wouldn’t it be better to do it differently? And very often the model’s answer was basically: “Yes, you are absolutely right.”
So the real challenge was not just generating code. The real challenge was making it solid and secure.
Because this is not a standard SaaS. It requires a lot of technical understanding. Not just coding itself, but understanding architecture, code concepts, data relationships, browser logic, desktop-specific things, and security principles.
After around 3 months of pushing aside almost everything else I had to do, and coding 10+ hours a day, I brought it to the level where my product is now competing with some of the biggest players in the industry.
Then I went even further and started automating SEO with OpenClaw.
Now it scrapes competitor articles, analyzes important keywords, adds them into our keyword database, and builds article drafts around that. It also generates a cover image with nanobanana-2 and does a lot of the repetitive work automatically. I have separate cron jobs researching new keywords through the Brave API. Another one pulls impressions from published articles, and when some topic starts getting attention, we duplicate the direction and create more content around similar keywords.
This is actually the part that changed my view the most.
These days, you do not have to be a developer to build things like this. But you do need vision. You need to know what you want to build, how it should work, and how to ask the right questions. And maybe the most important part — don’t rush.
Because with AI, rushing usually creates trash much faster.
Measure seven times, cut once.
Look at where we are right now.
The coding problem is almost solved. What matters more now is understanding what to build and how it should work. It’s becoming an era of product people, not just developers.
At the time I’m writing this, Opus 4.7 just came out. It can literally learn anything you need, explain it in multiple ways, never get tired of your questions, and go deep into research. It can analyze competitors, suggest ideas, help with marketing, and walk you through different scenarios.
This wasn’t possible before.
There has never been a better time to take an idea and actually turn it into something real.