Over the past few months, I’ve been experimenting with app generation tools like Lovable, Bubble, Base44, Claude Code, and ChatGPT. Internally, we’ve also been sharing our development cycle on an internal tool called ReadySetGo that generates Rails apps using thoughtbot’s best practices.
Like many people who are no longer in the code day-to-day, I was curious how quickly I could knock the rust off and build a real application with no-code tools. I’ve dabbled over the last year or so, but having a specific product in mind to build helped me better understand how much these tools have fundamentally changed the speed at which ideas can become businesses.
At thoughtbot, we’ve always moved quickly. Historically, a prototype might take us about two weeks to design and build. An MVP could take three or four months. Those timelines have worked well for years. But the tools we have today have compressed those cycles dramatically.
Recently, I built two small applications using Lovable. One is ShareCheck, which helps people check the credibility of links, pasted text, or screenshots before sharing them internally. The other is Hablayaya, a lightweight tool for practicing conversational Spanish with an agentic amigo.
Each was built in a weekend.
If you had told me a year ago that I could build two working apps so quickly, I would have assumed they were just demos. But these are real products that people can use.
The speed is exciting, but to be honest, it also feels a little reckless. While the barrier to building software has dropped dramatically, the barrier to building the right product has not. In fact, it’s probably higher than ever.
The new wave of AI-built apps
Many of the founders we’re seeing today already have what you might call a Lovable MVP. It works, it looks polished, and it proves the idea can exist. What they often don’t yet have is clarity about how to turn that prototype into a sustainable product and business.
In the past, product managers, designers, and engineers acted as natural gatekeepers. Features were debated. Ideas were tested before they were built. Sometimes that slowed things down, but it also forced teams to think carefully about what they were building.
And today, that friction is mostly gone. Anyone can open a no-code tool, describe an idea, and watch an application magically appear. Adding features can feel as easy as continuing a conversation with ChatGPT.
I completely understand how empowering that feels. It also makes it so incredibly easy to build the wrong thing very quickly.
When fast becomes messy
While building these apps, I ran into a pattern we’ve all experienced with ChatGPT. The system gets better and better as you iterate. Each change improves the product.
Then suddenly it starts to drift.
Something that worked perfectly an hour ago breaks. The logic becomes more complicated. Fixing one thing creates two new problems. It becomes surprisingly hard to get back to the simpler version you had earlier. For those who have trained LLMs, it may give you that feeling of overfitting.
The result is that AI-generated apps often accumulate unnecessary features and code complexity. The tools are fantastic at helping you build quickly with confidence, but they’re not particularly disciplined in helping you build simply and strategically.
And this is where product strategy comes in.
How to turn a Lovable prototype into a real product
If you’re experimenting with these app-generating tools like Lovable or Bubble, my advice is to lean into the speed. Use these tools to prototype aggressively. Try ideas that would have felt too expensive before. Develop your own opinions about the product. The goal at this stage is to learn fast while it’s cheap-ish.
You want to answer critical questions like:
- Who is this for and when will they use it?
- What problem does it actually solve?
- Which features will matter most, and which ones can wait?
Once you start feeling confident that it’s ready to bring to market, the challenge changes. It’s tempting to believe that you just need to deploy what you’ve built or apply a mobile wrapper to launch it, but in practice, that seldom works.
AI-generated apps often contain hidden complexity and unnecessary features. More importantly, the product usually reflects the assumptions of the person building it rather than the needs of the people who will use it.
When teams work with thoughtbot to launch a new product or business, we typically start with a Shaping Sprint that identifies the two or three biggest risks that could sink the business before it even gets started.
If we can identify those risks early, we can design the MVP around answering them quickly. Anyone can write good code. Many people can design beautiful interfaces. The hard part is making sure you’re not building the wrong thing.
Building MVPs like transportation
Many people imagine building a product like building a house. You start with the foundation, then the walls, then the roof. We often explain product development using a different metaphor, transportation.
You start with roller skates, and when you need to go faster, you can use the wheels to build a scooter. That scooter gets an engine and it becomes a motorcycle. You’re flying now but you need more room for cargo, so the engine is repurposed into a truck. And while you are getting to work in that truck, the team may already be building your dream jet in the hangar.
Even if someone handed you the perfect jet on day one, you may not yet have the skills to fly it.
Our job is to meet teams where they are in their journey to help them build the right mode of transportation for their needs: their budget, their destination, and their pilot ability.
Building software has never been easier
These tools are making it easier than ever to turn big ideas into working software. Prototypes can now appear in a day. MVPs can now be measured in weeks. While yes, this is a huge shift, the fundamentals of product development haven’t changed.
It’s more important than ever to understand your users. Assumptions must be tested while they’re still efficient to adjust. And most importantly, that you’re shipping the smallest version of the product that will deliver the most value. Yes, building software has never been easier. Building the right product is still the hard part.