I have a marketing degree I never used and I can't write production code. In the last year I built VibeBoost, a real X growth tool with auth, payments and a live paid API integration, in about two hours a day around a full-time job, faceless, from Lyon. I've also watched most of my own experiments make nothing, and learned more from the ones that broke than the ones that worked.
These are the patterns I'd tell a younger version of me. Not rules. A compass. If you've got a 28th, I want to hear it.
On the idea
1. Rebuild the one thing your competitor does best, first. Don't invent. Find the product people already pay for, find the single thing it does well, and rebuild that slice. The market has already validated it. You're just doing it sharper.
2. Pick a wave you didn't create. Build on a trend, a platform shift, a new rule people are already talking about. When X banned automated replies in early 2026, that ban did half my positioning for me. The wave markets for you.
3. If a competitor could copy-paste your landing page, your idea is too generic. Your edge should come from something only you know or lived. Mine is that I'm building from zero, in public, and I'll show the ugly numbers. Nobody can copy that.
4. Validate with a credit card, not a compliment. "Great idea" is worthless. The only validation is someone paying, or at least putting a card down for a trial. Friends will lie to be kind. Stripe won't.
On building with AI
5. Never prompt the whole app. Ask an AI builder for the entire product in one shot and you get spaghetti you can't fix. Build one screen. Confirm it works. Then the next. Slowness here is speed later.
6. Treat the AI like a fast junior dev you still have to direct. It's not magic and it's not your CTO. It executes what you describe. The quality of your description is now the entire job.
7. When it breaks something, don't re-prompt harder. Read the actual error. Tell it the one thing that's wrong. Make it fix only that. "It doesn't work" gets you vague, confident, wrong fixes.
8. The vision is no longer optional, it's the only thing left. AI removed the boilerplate, which means taste, judgment and knowing what to build are 100% of your value now. Spend your real hours there, not on setup.
9. Ship the smallest thing that delivers the core feeling. For me that's one reply drafted in your voice. Everything else is decoration around that one moment. Find your one moment and get a stranger to it fast.
10. Your first working version should embarrass you a little. If you're proud of v1, you waited too long. I shipped ugly and fixed in public.
On money (the part that almost killed me)
11. Margin is a feature. Build it before you have users. I wired up the X API, which charges per post read, and watched a $10 balance drain in two days because my app was re-reading data it already had. A product that loses money per user isn't a business, it's a hobby with a Stripe account.
12. Instrument your costs from day one. I couldn't fix my API bill until I logged every single call. You can't cut what you can't see. Add usage tracking before you add features.
13. Design against how you're billed, not how you assume you're billed. I assumed X charged per request. It charges per distinct post per day. That one wrong assumption was most of my waste. Read the pricing page like it's a contract, because it is.
14. Know your unit economics before you know your logo. I see indie founders agonize over a landing page while having no idea what one active user costs them. Cost per user first. Colors later.
15. Your own usage is not your customer's usage. I'm the app owner, so some of my API reads are cheaper than a real customer's. I almost priced off my own inflated-then-discounted bill. Model the customer, not yourself.
On pricing and the landing page
16. Sell the outcome, not the feature. Nobody wants "AI reply generation." They want to stop staring at a blank timeline. Features are vehicles. Sell the destination.
17. Use numbers, not adjectives. "Grow faster" is forgettable. "Your first 1,000 followers in 20 minutes a day" is a picture someone can hold. Specific beats impressive.
18. Describe the problem better than the customer can. People trust your solution only after they believe you understand their pain. I spend the top of my landing page on the freeze you feel opening X with nothing to post. Empathy before pitch.
19. One screen, one idea. Don't say everything at once. Each section earns the scroll to the next by making exactly one point. Confuse and you lose.
20. Make the pricing impossible to miss and impossible to overthink. Put pricing in the header. Offer three tiers, not seven. Every extra choice is another reason to close the tab.
21. Your CTA should say what happens next. "Get started" means nothing. "Start your first reply session" tells them exactly what they're about to do. Remove the uncertainty.
On distribution (where I'm weakest, and honest about it)
22. Building is the easy 20%. Distribution is the 80% that decides everything. I had a working product and six signups. That's not a product problem. Nobody knew it existed. Solve "who will see this" before you polish another feature.
23. The cheapest distribution for a founder is being a person people can see. Not a corporate feed. Your face, your voice, your real decisions. B2B buyers trust people over brands, especially early.
24. Share metrics and decisions, not vibes. "Building in public" fails when it's motivational quotes. It works when it's "here's my revenue, here's the mistake, here's what I changed." The ugly numbers are the content.
25. Own 3 to 5 topics for 90 days and post nothing else. Pick your lanes. Mine are building with no code, growing on X from zero, and the honest numbers. Consistency in a few lanes beats being interesting about everything.
On not quitting
26. The work is slow and boring and that's the filter. Three weeks of replying got me about 15 followers. Most people quit right there. The ones who win aren't smarter, they just showed up on the days it wasn't fun.
27. Watch your own behavior, it's your best product feedback. I sat a week not wanting to reply, then realized that's the exact wall my users hit. The thing you avoid is the thing to fix. Your churn is often just you, honestly.
These got me from "I quit every project" to a live product with paying users and a real business to fix. They won't all fit you. Steal the ones that do, and tell me your 28th.
FAQ
Can a non-technical person really build a SaaS? Yes, a real one with payments and integrations. You'll still hit genuine engineering problems, like API costs, that require thinking, not just prompting.
What's the single biggest mistake you made? Not watching my API costs until they were already leaking. Margin is a feature and I built it late.
How do you find time as a solo founder with a job? Two focused hours a day, one slice at a time, and refusing to touch the boilerplate AI can do for me.