Vibe coding with my 5-year-old: ⚡Bolt vs ♥️Lovable making an out-of-this-world space app

In many ways, vibe coding is like parenting young children. It can be magical, at times mind-blowing, and speckled with moments of joy.

Of course, it also comes with its share of repetition, frustration, and complete exaperation. One step forward, two steps back (and the occasional spoon thrown at your face).

Combining both? 🤯 After the last blog’s revelation that I’m cheap, this time I reveal that I’m a bit of a sadist.

My five-year-old is interested in Space 🌠 so I decided to challenge my two top prototyping tools, Bolt ⚡ & Lovable ♥️, to see who could build an interactive UI – with the champion likely to win my hard-earned $20 a month as I try to build out a full app with database, payments, and maybe LLM integrations.

The idea

SpacEd – an education app aimed at parents of young children. It would display fun space facts across different topic areas and have a mini-quiz at the end. It must be educational, easy to use (mobile friendly), and visually compelling. I planned to add login pages to track progress, so both have mocked dashboards.

My five-year-old has the final say who wins. But any opinions are welcome!

I used the same basic initial prompts and produced similar but noticeably different products. There was one common issue that I needed to fix manually before some minor cosmetic changes and we were ready to test the prototypes in the ‘market’ within a few hours!

Initial impressions

I have been playing around with these tools (+vO) at work, and iI’ve been getting a feel for their strengths and weaknesses; these came through here as expected (vO has a less generous free tier, so I did not include it).

I have found Lovable slightly quicker, and usually prettier though with more (self-detected) bugs. Bolt is a little less expressive, reports fewer bugs itself (either because there simply are fewer, or they are not called out) and feels a little more ‘solid’ or well thought out. I should emphasize these two tools have far more similarities than differences – I mention that Lovable is often prettier yet some of the detailed touches added in the Bolt prototype were brilliant, the storms on Jupiter and Neptune especially jumped out to me. Aspects of both came out on top.

Head to Head

I used a basic PRD that I sent in prompt 1. I used the same follow-up prompt “Let’s improve the graphics on the home page – show all 8 planets and the sun – add a comet passing by” and after that moved to fix the independent issues to get to launch 🚀.

♥️ Lovable – https://star-facts-galaxy-fun.lovable.app

What went well:

• Strong V1 output. It’s beautiful right away, polished with some great touches.
• Well-designed: delivered the 3 topics, with 5 facts each, and a 3-question quiz.
• Appealing use of colors. Nice details – for example, the stars in the background.
• Custom request (for a comet) was nicely fulfilled in one-shot.
• Age-appropriate facts (good for 5-year-olds).

What needed work:

• Initial homepage had only 3 orbiting planets — later corrected to 8, but the orbits were off.
• Fact images failed on the first attempt (only 1 unrelated image). On retry, images were added per fact but they were not all space-related — manual image curation needed.
• Slight confusion around login vs. guest mode — possibly due to unclear prompting.

Prompt usage notes:

• Spent prompts on:
  • Fixing broken image links – in the end, I needed to add images directly in a file on GitHub.
  • Creating a new login page – this went well.
  • Resizing images and adjusting text colour (minor tweaks).
  • Fixing the solar system graphic, from Saturn’s rings to Uranus’ orbit (wow, if these were not free I would have been frustrated!🤯)
• Other notes:
  • System located 2 bugs and fixed them itself (I see this often with Lovable).
  • I used lots of “chat” before asking for changes, this helped, it was free the weekend I was testing.
  • I mostly used Anthropic to good effect, though some refactorings produced weird results and it took a switch to ChatGPT for it to immediately see an issue Claude missed 10+ times.
  • This pushed the build time to ~2 hours.

---

⚡ Bolt – https://bolt-spacded-app.netlify.app/

What went well:

• Also strong, well-designed V1. It is not quite as beautiful, but feels more solid. Nailed login page.
• Good 5-year-old-friendly facts (similar to Lovable — likely using same LLM/data sources).
• One image per fact — correctly followed instructions but not all were relevant!
• 5-question quiz.
• Planets had fabulous detail – particularly the storm spots on Jupiter & Neptune.
• The orbits were staggered, slower the further out, nice.

What needed work:

• Most images were not on-theme.
• Homepage was more basic compared to Lovable – though logo was better.
• Not all planets were added initially.
• Quiz was created correctly but rendered all in white, so was not visible – fixed in 1 prompt.
• Comet, despite several attempts to improve, still not as good as Lovable’s version.☄️

Prompt usage notes:

• Spent prompts on:
  • Fixing broken image generation – in the end, I needed to edit URLs manually in-code.
  • Fixing Quiz display.
  • Improving comet (it improved, but is not great) & adding all 8 planets (sorry Pluto!).
  • Resizing images (minor tweaks).
• Other notes:
  • System located no bugs itself.
  • I used “chat” here too before asking for changes, this is a great feature.
  • Build time under ~1 hour, fewer prompts.

Feature

Lovable ♥️

Bolt ⚡

Initial design quality

Good, pretty, excellent colours

Solid, all requested pages built, less polished visually

Mobile display

Looked good

Looked good (has views per phone type – nice!)

Image request

Single images initially, manual fixes needed

More relevant images, but manual URL edits required

Home page visuals

Pretty, hard to fix orbital inaccuracies

Not as pretty but easier to improve, great details

Comet graphic

Excellent, one-shot fulfillment

Not great after multiple attempts

Bugs and errors

Self-reported and fixed two bugs, more overall difficulties

Fewer bugs overall, none self-reported, quick fixes

Build time

~2 hours

~1 hour

Common Issues

Both struggled with adding images – they have different default image websites and I realized the issue is there; the free versions have limmited images and the models are not the best at seeing just yet. So this is more a requirement issue my side, I could pay up for a better tool or pursue an alternative solution – I did try using graphics over pictures but they were a little too rudimentary so I reverted back. The ability in both tools to revert to the previous build is brilliant (and essential); it allows you to try big ideas, to not worry about your baseline when you ask for unpredictable changes.

Correcting the 8 planets, their orbits, and Saturn’s rings really caused issues on Lovable. I had a nightmare with Claude, with poor Uranus being stuck behind the sun 🌞 ~10 edits did not fix this – ChatGPT located the issue immediately: “I found the specific issue! Looking at the planet data, I can see that Uranus has an orbital size of w-84 h-84. The problem is that Tailwind CSS doesn’t have a built-in w-84 or h-84 class.” Oh obviously, all PMs know that 😂

Seriously though, having the ability to switch models within Lovable was an awesome one-weekend-only feature; make it full-time please!

Product Feedback ⚡& ♥️

Prototyping 🏗️ – Amazing for getting you to 80-90% done – you can often do this in just a few prompts. It is trivial to knock out 5 variations on a new prototype inside an hour – all without developer or design assistance. This is a game-changer for PMs, a force multiplier.

MVPs – The final 20% or so… harder. More painful. You can do simple MVPs without much stress. More difficult ones too with more time & experience (+ credits!).

Full independent Apps – harder still. The final 5%, the last complex error or the thing that makes you stand out in a crowd? Designer & Developer skills would certainly be appreciated!

Integrated Apps – the hardest; here, detailed knowledge of the existing stack is needed to power integrations. You will struggle to prompt your way out of a P0 on either tool.

If, as some predict, Product, Engineering, and Design converge, it is through tools like this that real value will emerge – leveraging existing skills and supplementing areas of weakness. PMs and designers would benefit from grounding in code, or maybe more interestingly, tasteful, customer-focused Engineers can truly excel.

Next steps

I was able to easily get prototypes up and running; building a full app will be harder but I hope doable. This is for fun, but a key consideration for me would be to make this in some way better than a book as I’d rather have my five-year-old reading 📖 compared to using a screen. Here, an LLM integration may make all the difference. More to come!

Results

The five-year-old went for Lovable (though I’m not sure he understood the difference). For me, Lovable wins home screen and progress page, Bolt the quiz section and general vibes on build quiality. Family members were split evenly, tell me what you think!

♥️ https://star-facts-galaxy-fun.lovable.app ⚡ https://bolt-spacded-app.netlify.app/