Lessons from Lovable

Both apps were built almost entirely on a phone since one hand was occupied feeding or walking around with my newborn son. Voice memos into Claude initially led to a structured requirements doc in Google Drive. I also planned build phases with Claude before touching Lovable. Batching changes using Claude to write the prompts saved credits and produced better output than one-at-a-time prompting in Lovable. 

Claude was also my debugging partner: writing SQL queries to paste into Supabase. In the World Cup Sweepstakes app, this included diagnosing live match status bugs, tracking a silent data-wipe bug caused by a blanket upsert overwriting manual card data every 5 minutes.

I just needed to understand the problem clearly enough to explain it.

These tools and this tech is so impressive. I loved the workflow. It gets you very far, very fast. The focus was on the UX. The initial pass at the UI for both apps was ugly and tried too hard, but one detailed prompt for a clean, minimal UI was enough to get close to the right aesthetic from the start.

The gap between the idea and a working prototype you can actually feel and use is almost nothing. 

The iterative loop: Prompt it, use it, identify gaps, fix it, repeat. Fast, fun and extremely satisfying. 

My own experience is pretty much in line with the general consensus: 

• AI app builders are excellent for personal projects and prototyping. Not finished product quality without serious refinement.
• They get you to ~80–90% fast. The last 10–20% is where design skill matters most.
• Anyone can get to 80%. Getting the last 20% right is the craft.

Some of the drawbacks: 

• The lack of granular UI control is frustrating. Credits spent on small visual tweaks felt expensive. 
• Component reuse is a problem (e.g. same segmented control built twice as separate components)
• Presumably the underlying code quality is untidy. 

It’s clear my role in the industry is changing, but I feel more valuable than ever.

Playing with Lovable built my confidence in the value of design skills. I feel like when the web was built only by developers, it took some time to realise there was a massive space for designers to step in and occupy and improve things. More time revealed that we could benefit from specialist designers in UX and UI, both of which are often broken down even further. Now, with these powerful tools, individuals, designers or developers, can achieve so much more. It feels to me like the pendulum is almost swinging back again. With AI at their side, an individual with a broad profile; UX instinct, UI sensibility, and a surface technical understanding, can build today what would have needed a full team of specialists only a couple of years ago. 

The output might need a technical review but I’m talking from the design point of view here and it’s very exciting.