AI Creator Interview with @Kappaemme1926 → Today I spoke with Francesco — a 21-year-old indie hacker from Italy who builds products, posts on X, and uses AI not as a trend but as a tool to really speed things up In the interview, we discussed: ▸ why he trusts Codex the most ▸ why he thinks Claude is overrated ▸ how he’s building Slashcmd —a marketplace for AI agent skills ▸ what he doesn’t want to automate in his work ▸ how AI helps him run X ▸ what he would do with $100 and zero audience in 30 days The full interview is below ↓ › Rexei: Introduce yourself, tell us a little about yourself and what you do › Kappaemme: Hi, I’m Francesco. I’m 21, Italian, a computer science student, and an indie hacker building in public on X. I’m focused on AI, coding, and creating useful products from scratch. I use X to share what I’m learning, what I’m building, and the real process behind it: wins, mistakes, experiments, and lessons. My goal is simple: use AI not just as hype, but as a practical tool to build faster, automate boring work, and turn ideas into real products. ⸻ › Rexei: What does your toolkit look like right now (Claude Code, Codex, Cursor)? Which service or model have you come to rely on for effective work? › Kappaemme: Right now, the tool I use the most is Codex. I rely on it a lot for coding, debugging, understanding projects, and moving faster from idea to execution. I really like how OpenAI designed the product. It feels practical, clean, and focused on real development work, not just on generating random snippets of code. For me, Codex is not only a coding assistant, but more like a technical partner that helps me think through problems, build faster, and improve my workflow. I’m aware of tools like Claude Code and Cursor, and I think they’re interesting, but right now Codex is the one I trust the most and the one that fits best with how I work. ⸻ › Rexei: Which AI tool do you think is the most overrated? And which one is the most underrated? › Kappaemme: I’d say the most overrated AI tool, for my personal workflow, is Claude. I know a lot of people love it, and I understand why, but it has never really clicked with me. I don’t like Anthropic’s approach as much, and when I use Claude I don’t feel the same level of control, speed, and practical value that I get from OpenAI tools. It’s not that Claude is bad. I just think it gets praised a lot, while for the way I build, code, and think through projects, it doesn’t fit me that well. The most underrated tool, in my opinion, is Cursor. Not because nobody knows it, but because people often focus only on the model battle, like GPT vs Claude, and forget how important the actual workflow is. Cursor is powerful because it brings AI directly into the coding environment. When used well, it can make the whole building process faster, smoother, and more natural. For indie hackers and builders, that kind of workflow advantage matters a lot. ⸻ › Rexei: What are you creating now and why did you choose this particular idea and not a hundred others? › Kappaemme: Right now, I’m working on an application for a personal trainer, for a real client. It’s a more practical client project, and it’s helping me improve the way I build products around real needs, not just ideas. But my latest personal project, and probably the one that represents me the most right now, is Slashcmd: a marketplace for AI agent skills. I chose this idea because I experienced the problem myself. There was a period where I started using skills a lot to optimize my workflow, especially when working with AI tools. I realized how powerful they can be when they are reusable, easy to install, and connected directly to the way builders actually work. So instead of choosing a random idea from a list of a hundred possibilities, I chose something that came from my own workflow. I had already felt the value of it personally, and that made the idea much easier to believe in. For me, the best projects usually start like that: you solve a problem you actually have, then you try to make the solution useful for other people too. ⸻ › Rexei: What part of your work process is already fully automated, and what do you fundamentally not want to automate yet? › Kappaemme: I don’t think my process is fully automated, and honestly I don’t want it to be. The repetitive parts are the ones I automate or heavily accelerate with AI: boilerplate code, debugging, first drafts, project understanding, small implementation tasks, and workflow optimization. Codex and skills help me remove a lot of friction and move faster from idea to execution. But I don’t want to automate the parts that require taste and judgment: choosing what to build, understanding users, making product decisions, and deciding if an idea is actually worth pursuing. For me, AI should make me faster, not make me passive. ⸻ › Rexei: What's the biggest mistake newbies make when they start creating something with AI? › Kappaemme: The biggest mistake is using AI passively. A lot of beginners write vague prompts, give no context, and expect perfect results. But AI works much better when you give it clear instructions, constraints, examples, and a specific goal. The better the prompt, the better the output. Vibe coding is great, but it has to be done with a direction. You still need to understand what you’re building, check the code, test it, and make decisions. AI should accelerate you, not think for you. ⸻ › Rexei: Let's talk a little about your journey at X: How did your journey at X begin? Did you immediately know what you wanted to write about, or did you only find your theme through experimentation? › Kappaemme: My journey on X started very organically. I didn’t have a perfect strategy at the beginning. I was just a tech fan sharing my thoughts, the products I was building, and my take on what was happening in the tech world. A big reason I started was my friend Emanuele. He inspired me and helped bring me into this space, and that had a big impact on where I am now. At first, I experimented a lot. I wrote about tech news, AI, coding, products, and whatever I was learning. Then, little by little, I found my theme: building in public, AI, indie hacking, and sharing the real process behind creating products. I think people started following me because the content was genuine. I wasn’t trying to look perfect. I was sharing what I was learning, what I was building, and how I was evolving. ⸻ › Rexei: How does AI help you run X: search for ideas, analyze your audience, write drafts, plan content, or do you prefer to do everything manually? › Kappaemme: AI helps me with almost everything on X: brainstorming ideas, finding angles, organizing thoughts, writing drafts, improving hooks, and planning content. Usually, when I have a project or something I want to share, I start by talking with AI. I use it almost like a thinking partner: I exchange ideas, test directions, and turn messy thoughts into a clear plan. But the final voice is still mine. I don’t want AI to make my content generic. My opinions, my taste, and my experience are what make the content real. I honestly don’t think I could go back. Manual work still exists, but it has changed. Now I spend less time fighting the blank page and more time deciding, editing, and making the content feel authentic. ⸻ › Rexei: If you had to start from scratch: a $100 budget, no audience, but AI tools. What would you do in the first 30 days? › Kappaemme: If I had to start from scratch with a $100 budget, no audience, but AI tools, I wouldn’t try to build a huge product immediately. In the first 30 days, my goal would be simple: find a real problem, build something small around it, and start creating content around the process. During the first week, I would choose a clear niche, probably around AI, builders, or productivity, and use AI to research problems, brainstorm ideas, analyze what people are already talking about, and create a simple content plan. Then I would start posting every day on X. Not polished content, but real content: what I’m learning, what I’m building, mistakes, small wins, and useful insights. At the same time, I would talk to people, reply a lot, and try to understand what problems keep coming back. In the second half of the month, I would use Codex and AI tools to build a very small MVP around one of those problems. Nothing too complex. Just something useful enough that people can try, give feedback on, or even pay a small amount for. I would spend the $100 only on essentials: a domain, basic hosting, maybe one or two tools that help me move faster. I wouldn’t waste it on ads at the beginning, because with no audience and no validation, paid traffic can disappear very quickly. By the end of the 30 days, my goal wouldn’t be to have a perfect business. My goal would be to have three things: a small audience that knows what I’m building, a real problem I understand better, and a simple product that I can improve from feedback. → Thanks to Kappaemme for joining this interview Let us know who you'd like to see next in the interview See you later

Read the post on X