Frontly: Building an AI App Builder Before the Market Existed
Frontly started in 2021 as a straightforward no-code app builder. Nothing AI about it. Just a platform for building custom business applications without writing code.
Then ChatGPT launched in late November 2022, and everything changed.
I bought the frontly.ai domain and quickly shipped a new product: a platform where you could create reusable AI prompt templates with variables that appeared as simple forms. Your team fills out the form, the prompt runs behind it, and they get consistent AI-generated output without needing to write prompts themselves. No one was doing this yet. ChatGPT had just arrived and most people were still figuring out what it was.
But the prompt template product was just the starting point. Within weeks I realized the bigger opportunity was back in the core app builder. If I could let users add an “AI action” as a step in their applications, they could build tools that actually use AI - not just generate text, but integrate AI into real business workflows. So I went back to the platform and started building that.
By February 2023, ChatGPT was integrated directly into the builder. By April, I was demoing full AI app generation from a single prompt: describe what you want, and Frontly generates the pages, components, data models, and layouts. The generated apps were simple, but for early 2023, it was genuinely impressive. This was over a year before Lovable, Bolt, or “vibe coding” became mainstream concepts.
What people built
Frontly powered a lot of internal business tools. Custom data management systems, CRM-style platforms, workflow tools - the kind of applications where every business has slightly different needs and off-the-shelf software never quite fits. A lot of our best users were agencies building custom SaaS and internal tools for their clients, using Frontly to move fast without hiring developers.
None of it was flashy. But that was the point. These were real tools solving real operational problems for real businesses, and many of them are still running in production today.
Growth
As a technical solo founder without much funding, I didn’t have a marketing team or a growth budget. I focused on the product, talked to customers daily, shipped fast, and used launches on AppSumo and Product Hunt to find users. That approach worked. 20,000+ monthly active users. $600K in sales. #1 Product of the Day on Product Hunt.
The users who stuck around validated the core idea: there’s a massive market of people who need custom applications but can’t or won’t write code, and AI can dramatically lower the barrier to building them.
The evolution
The platform went through several major phases, each one building on the last:
2021-2022: No-code app builder. A basic platform for building custom business applications. No AI, just drag-and-drop application building.
Dec 2022: AI prompt templates. Launched on frontly.ai. Reusable prompt templates with variables, presented as forms. Your team uses AI through consistent, tested prompts without needing to understand prompt engineering.
Early 2023: AI-enabled apps. Integrated ChatGPT into the core builder. Users could add AI actions to their applications, making the tools they built actually intelligent.
Mid 2023: AI app generation. Type a prompt, get an app. The generated apps were simple, but this was the first version of what the industry would later call “vibe coding.”
Early 2025: V2 vibe coding platform. A full rebuild with a more agentic experience. The AI generation pipeline now used a multi-step process with human review at each stage - the architecture I described in Chaining AI Prompts Before Agents Existed. Each generation step would pause, show the user what it created, let them edit and approve, then continue building on that validated output.
That last evolution - the multi-step pipeline with human-in-the-loop - was the most technically interesting work I did at Frontly, and it directly informed everything I’ve built since.
What happened
I raised $400K from investors including Andrew Wilkinson (Tiny Capital), which should have accelerated things. Instead, it added pressure to chase traction instead of going deep with a specific customer segment. I was building broadly when I should have been building specifically.
Then, in late 2024 and early 2025, the market exploded. Lovable and Bolt took off with dedicated teams and venture backing. Google launched AI app building tools. Replit shipped its agent. Even Figma launched an AI app builder. The space I’d helped pioneer was suddenly one of the most competitive markets in tech, and every new entrant had more resources than I did as a solo founder.
I’d been building AI-generated applications for over a year before any of them launched. But being early doesn’t matter if you can’t defend the position. And I knew, watching the wave of well-funded competitors arrive, that the window for Frontly to become a venture-scale company in this space had closed.
What I took from it
Frontly taught me two things that shaped everything I’ve built since.
First, early entry without focus is just a head start you waste. If I’d locked in on a narrow customer segment from day one - agencies building client tools, for example - I’d have built relationships and product depth that would have been hard for well-funded latecomers to replicate. Instead, I built broadly and got outscaled. The lesson wasn’t about timing. It was about specificity.
Second, AI does dramatically better with small, focused steps and human oversight than with single-shot generation. That insight came directly from building Frontly’s generation pipeline. Single-prompt app generation produced garbage. Multi-step generation with human review at each stage produced applications people actually used.