Hello, I'm a freelance startup programmer in Korea, and my name is Jeong Dong-woo.I share many of the same concerns as you.I've been on an unconventional career path for about 7 years. My specialty is building factory equipment control programs using WPF on the frontend with Ladder Diagram logic as the backend. In this field, ladder logic interpretation is still weak for AI, and because of the security restrictions at Korean factories AI tools are difficult to use on site, so I've been able to keep making a living here.My typical work involves building PLC-based control programs with WPF and integrating them with MES systems, and this is my core specialization. The downside is that because of the labor-intensive nature of this work, scaling up is genuinely difficult. In practice, it feels very much like being a subcontractor to the actual equipment vendors.I started out on a freelance community platform, and for the first 3 years I had almost no work, so I took on manual labor in parallel. Currently I'm delivering code to Hyundai Motors as a second-tier subcontractor, but I still have many of the same worries you describe.Even calling it a business, I only earn around 4 to 5 million Korean won per month, and I honestly don't know how to grow it beyond this. Reading your posts, I feel reassured to see that many people are wrestling with similar concerns.To try to expand, I've also built and delivered SaaS products. My typical engagement is a solution of around 60,000 lines of code. (Over 7 years, I've completed roughly 40 corporate outsourcing projects.)Every product I launched directly to the market has failed. The only things that worked for me were building SaaS for companies that already had revenue, and making industrial control programs.Reading the advice posted here, I keep coming back to the same realistic conclusion: in the current AI era, because of the AI SLOP problem, no matter how good a program you build, it simply doesn't reach users. That's the core issue. I've been increasingly convinced that flooding the market with many AI-generated programs is actually a more effective strategy than carefully crafting one good one. When I look at the programs that are generating real revenue recently, they seem to follow this pattern.I no longer believe that a good program necessarily succeeds.In the end, the key is exposure, and this is where I keep hitting a wall. No matter how much I invest in SEO, my Korean website struggles to gain visibility. Right now I'm using an AI API to translate my Korean articles into English and posting them as a form of promotion, though I'm not yet sure whether this approach will work.Paths to success differ from person to person and from country to country, which makes specific advice difficult to give. But in my view, what matters most is how much initial capital you can put into Google ads.Expensive ad inventory is generally personalized. Users who don't provide much information to Google get served low-tier ads instead. So my current thinking is the opposite of the conventional approach: place a large volume of low-tier ads, and build programs that appeal to users in the "internet-underserved" demographic, meaning people who don't hand over their information to the major platforms.That said, I don't want to go down an illegal path, which is why I'm not building dating apps (in Korea, dating apps often end up tied to drug dealing or prostitution, so I'd rather stay away from that space).Unlike many of the other comments here, I don't believe that a good program necessarily succeeds. The real question, in my view, is how to gain direct access to consumers.