Latest / The Indie Hacker Podcast with Fexingo: Solo Developers, SaaS Side Projects, and Independent Tech / How a Solo Dev Built a SaaS With Zero Code Using AI Agents
Transcript
- Lucas: Alright, so I want to talk about something that feels almost too good to be true — but apparently it's working for a growing number of solo developers. The idea that you can build a functioning SaaS product without writing a single line of code yourself, using AI agents. Luna: Zero code? Like, you literally just prompt your way to a product? Lucas: That's the claim. And I found a concrete example that backs it up. A solo developer named Maria Chen — she's a former product manager, not an engineer — launched a SaaS product called FormLens in February of this year. It's a form analytics tool that tracks user behavior on web forms. And she built the entire thing using AI coding agents: Cursor, Replit Agent, and Bolt. Luna: Wait — three different AI tools? Why not just stick with one? Lucas: Great question. She used them for different stages. Cursor for the initial prototype and backend logic — it's good for longer, structured coding sessions. Replit Agent for quick iterations on the frontend and testing. And Bolt for rapid UI adjustments and debugging. She said each tool has its strengths, and she'd switch depending on what she was stuck on. Luna: So she's essentially orchestrating these AI agents like a project manager? That's kind of the dream for a non-coder. Lucas: Exactly. And here's the timeline: idea to a working MVP in three days. First paying customer in week two. By week fourteen, she hit five thousand dollars monthly recurring revenue. All with no technical co-founder, no prior coding experience. Luna: Okay, that's impressive. But I have to ask — what's the catch? Because I've tried using AI to build little scripts and it often gets stuck on edge cases or hallucinates libraries that don't exist. Lucas: You're right. She documented her process publicly on her blog, and she ran into exactly those issues. The AI would sometimes generate code that referenced a fictional API, or produce a component that looked right but had hidden logic errors. Her biggest time sink was debugging ai generated code. She estimates she spent about forty percent of her total time fixing things the AI got wrong. Luna: Forty percent? That's huge. So it's not 'zero code' in the sense of zero effort. Lucas: No, it's more like 'zero code from scratch.' She still had to understand enough to spot when the AI was going off the rails. But she didn't need to know syntax or algorithms. She needed to know what a form analytics tool should do, how the data should flow, what a user expects. That's product thinking, not software engineering. Luna: So the barrier shifts from 'can I code it?' to 'can I describe it clearly enough?' Lucas: Exactly. And that's a huge shift for the indie hacker space. Historically, solo founders who weren't technical had to either learn to code, which takes months or years, or partner with a developer, which often leads to equity splits and misaligned incentives. Now, Maria's story suggests a third path. Luna: But is this sustainable? I mean, AI models improve fast, but the tools she used — Cursor, Replit Agent, Bolt — they change constantly. What works today might break in six months. Lucas: She actually addressed that. Her strategy was to treat the AI tools as a moving target. She didn't commit to any one tool. She learned the general principles of prompting, testing, and iterating, and she's ready to switch as new tools emerge. She's already evaluating a newer agent called 'Devin' for version two of FormLens. Luna: That's smart — not getting locked into a specific platform. So what did she actually build? Walk me through FormLens. Lucas: It's a lightweight JavaScript snippet you embed in any web form — contact forms, sign-up forms, checkout forms. It tracks things like field abandonment, time spent on each field, validation errors, and completion rates. Then it surfaces insights like 'users are dropping off at the email field' or 'the phone number field takes seven seconds on average.' Luna: That's a pretty crowded space. Google Analytics does some of that, and there are tools like Hotjar. What's her differentiator? Lucas: Price and simplicity. She targets solo creators and small e-commerce stores. Her pricing starts at nine dollars a month for up to ten thousand form submissions. Hotjar's equivalent plan is thirty-nine dollars. And she keeps the UI dead simple — one dashboard, three reports. No heatmaps, no session recordings, just form analytics. Luna: So she's going after the underserved bottom of the market. Classic indie hacker move. Lucas: Exactly. And her early customers found her through a Product Hunt launch and a few niche communities like Indie Hackers and some no-code Facebook groups. She didn't spend a dime on ads. Her biggest marketing channel was being transparent about building with AI — people were curious and shared her story. Luna: That transparency is interesting. Do you think being open about using AI helped or hurt her credibility? Lucas: In her case, it helped. She framed it as 'I'm not a developer, but I can still build this thanks to AI.' That resonated with other non-technical founders who felt locked out. It became part of her brand. Some potential customers might worry about reliability, but she mitigated that by offering a generous free tier — up to a thousand submissions a month — and being responsive on support. Luna: So she's building trust through transparency and low risk. That's a solid playbook. Lucas: Yeah. And there's a bigger takeaway here. Maria's story isn't just about one tool or one person. It's a signal that the cost of building software is dropping dramatically. If AI agents can handle the coding, the real value shifts to understanding the customer problem deeply and communicating that clearly to the AI. Luna: Does that worry you at all? I mean, if everyone can build software, won't the market get flooded with mediocre products? Lucas: It's a valid concern. But I think we're already in that world — low-code tools like Bubble and Webflow have been around for years, and they haven't caused a flood of great products. The limiting factor isn't the ability to build; it's the ability to solve a real problem and distribute the solution. AI lowers the build barrier, but it doesn't lower the distribution or product-market fit barrier. Luna: Fair point. And Maria seems to have found that fit. So what's next for her? Is she planning to stay solo or raise money? Lucas: She's committed to bootstrapping. She wrote that her goal is to reach twenty thousand MRR by the end of the year without taking any outside capital. She's hiring a part-time customer support person, but she wants to remain the sole product decision-maker. She feels that staying lean forces her to stay close to customers. Luna: That's the indie hacker ethos in a nutshell. And it's refreshing to see someone who could have easily raised on a 'no-code AI builder' narrative choose not to. Lucas: Yeah. And speaking of narratives — if today's tech conversation gave you something usable, I want to mention something quick. We deliberately don't run ads on these shows. It's a choice we made early on because we hate the way ads interrupt the flow of a good conversation. Luna: Yeah, we've always felt that if the content is valuable, listeners will support it on their own terms. No forced breaks. Lucas: Exactly. So if you want to support that ad-free approach, the link is buy me a coffee dot com slash fexingo. That's it — no pressure, just a way to keep this sustainable for those who find it useful. Luna: And we really appreciate the support from those who do. Alright, back to Maria — she also shared some specific mistakes she made. What were the biggest ones? Lucas: The biggest was over-relying on AI for security. She initially used an AI agent to write her authentication flow, and it created a vulnerability where users could access other users' data by simply guessing a user ID. A beta tester found it and reported it. She had to rewrite the entire auth system manually — well, with heavy AI assistance — but she learned to never trust AI with security-sensitive code without thorough review. Luna: Ouch. That could have been a disaster if a malicious actor found it first. Lucas: Exactly. She also struggled with database schema design. The AI would create a schema that worked for the current feature but wasn't scalable. She had to refactor the database three times in the first month. She eventually learned to design the schema herself using a whiteboard before prompting the AI. Luna: So AI is great at execution, but not at architecture. That seems like a pattern we'll see more of. Lucas: For sure. And her advice to other non-technical founders is: learn enough to be dangerous — understand the basics of data modeling, APIs, and security, but let AI handle the syntax and boilerplate. That's the sweet spot. Luna: That's a good note to end on. So what's your takeaway for our listeners? If someone's listening and thinking about building something with AI agents, what should they do? Lucas: I'd say start with a very narrow problem. Don't try to build the next Salesforce. Pick one feature, one workflow, and use AI to build a prototype in a weekend. Then show it to real potential users. If they don't care, you've only lost a weekend. If they do, you've got validation and a foundation. That's exactly what Maria did with FormLens — she started with just 'show me where people drop off on my contact form.' Now she's at five thousand MRR.