AI Coding for All
Building 2 websites with Lovable
For this week’s post, we wanted to walk through a practical example of using AI Agents—specifically Coding Agents.
We chose Lovable because… well, we love it! But we’ll be exploring other tools in the future to compare results and share insights.
We tested two very different use cases:
1️⃣ Theo, my 13-year-old son, built a Latin revision website—in just 30 minutes!
📚 Check it out here
2️⃣ We built a household task organizer to replace chaotic WhatsApp threads between parents and nannies.
🏠 Check it out here
These experiments showcase both the power and limitations of AI-driven coding tools.
1️⃣ Theo’s Journey: Building a Latin Revision Website
Theo worked with Lovable to build study websites, primarily for Latin and Chemistry. An unusual combination, but Lovable didn’t mind that.
The Latin Site: A Surprisingly Smooth Process
He started with a pretty decent prompt (grammar mistake included!):
“Designing a Common Entrance Latin Revision Website: Content & Resources Guide Overview: A successful revision website for Common Entrance Latin (typically the 13+ exam) should mirror the exam syllabus and structure. Below is a comprehensive outline of content and resources to include, organized into key sections with suggested layout and navigation. This ensures students can easily find vocabulary lists, grammar references, exercises, past papers, exam tips, and cultural context relevant to the UK Common Entrance Latin syllabus.”
Then, he provided Lovable with a vocab list, and in return, he got a well-structured study site with:
✅ Organized topics
✅ Grammar tips
✅ Interactive study buttons
✅ Past paper questions with hints and model answers
Most features worked as expected, and Lovable was quick to fix issues when asked.
The Chemistry Site: Not as Smooth
Theo also wanted to build a science revision site with quizzes, grading, and interactive features.
While Lovable set up the basics, several technical issues arose:
⚠️ Broken buttons
⚠️ Error messages
⚠️ Unfinished features
Despite multiple attempts to fix the problems, the Chemistry site never fully came together—causing Theo some frustration with the tool’s limitations.
Key Takeaways from Theo’s Experience
🔹 Success isn’t always guaranteed—some projects will run smoothly, while others hit unexpected roadblocks.
🔹 Clear requirements help, but AI isn’t perfect—Lovable was responsive, but some requests were too complex for instant implementation.
🔹 Trial, error, and persistence matter—building something new, even with AI, isn't friction-free.
Or, in a nutshell:
Even with zero coding experience, a 13-year-old can create something meaningful using simple prompts—which is incredible!
Theo’s main lesson? Start simple. Go deep. The Latin site worked because it was focused and structured, while the Chemistry site struggled due to complexity creep.
Ultimately, the learning process is just as valuable as the final product.
2️⃣ Building a Household Task Organizer with Lovable
Step 1: Crafting the Right Prompt
We asked Perplexity to help craft a strong prompt: 💡 “I want to create a Trello-like tool for managing household errands with my nanny.”
Step 2: Perplexity’s Detailed Response
The AI-generated prompt included:
✅ Defined goals
✅ Key features (user roles, task stages, UI, future-proofing)
✅ Optional add-ons that were a pleasant surprise!
Task due dates & reminders
File/image attachments
Color labels or tags
Activity log
User assignments
Step 3: Iterating with Lovable
Lovable built the first version, and we refined it by requesting additional features:
🎉 A more engaging name → "Family Command Center"
📅 Due dates for tasks
🧑 Avatars to show who’s responsible
📆 Google Calendar integration
And this is what we got to:
Step 4: Key Issues Identified
While the experience was overall great, two challenges stood out:
🔹 Privacy Concerns → Right now, anyone with the link can access household task details, which raises some concerns.
Lovable suggested implementing user authentication via Supabase to improve security. However, the implementation is complex and lacks user access management (such as approving or revoking access).
🔹 Task Notifications → The site only checks for overdue tasks when it’s running, meaning you won’t know about overdue tasks until you open the site. Ideally users should receive proactive notifications for overdue tasks. However email alerts for overdue tasks require backend integration, which Lovable does not yet support.
Final Thoughts: What We Learned
This was a huge eye-opener—and a glimpse into where things are headed with AI-powered development.
Key Trends We’re Seeing
🚀 Hyper-personalization in building: users will build their own tools instead of relying on generic apps, tailoring tools and workflows to their preferences.
📚 Learning apps need to evolve: simple flashcards won't cut it anymore; interactivity will be key.
🔧 AI Agents will get more powerful: eventually, tools like Lovable will support backend automation (like sending overdue task notifications) natively
🔐 Privacy will be a major concern: and solving it won’t be easy
Next Up: Building Our Own AI Agent!
Now, it’s time to take everything we’ve learned and build our first AI agent.
Stay tuned!
As always..if you have any questions, insights, or fascinating research to share, we’d love to hear from you!
Tools we used:
Lovable
Grok
Our previous posts: