Lovable or Leavable?

I decided to participate in the free for all AI "vibe coding" weekend offered by Lovable. Of course, my perspective is one from a test and performance (always). So review, in case anyone cares:

I believe this was a customer driven load testing exercise for the back end engineers to see what their systems could handle. They were probably needing to do that anyway to determine scalability, but they took notes from Amazon and created their own "Prime Day". It turns into a marketing event that everyone talks about, and they don't have to go through any long and tedious end-to-end performance testing. Just test it live in production. In the end, they are probably going to get new customers out of it, so it's actually a win.

I've seen people raving about their experience with lovable, in spite of the problems with being overloaded with traffic. I spent the whole two days working with it to see what it would do. From the very beginning, I was getting "we are experiencing heavy demand." Then the constant queing started. Everytime you tried to submit, you would get a 60 second "wait to get back in the queue - or upgrade and jump to the head of the line" messages. At some points, you would hit the submit button 8 times in a row before the 60 second timer went away.

Anthropic is the clear winner in the showdown, producing the best code. They were the ONLY one of the models I saw that had a "rate limited" message when you looked at switching models. I think most people that know are going to choose them if given the chance.

If this were any other business, the end customer would have left the store and never came back. The experience from just a reality stand point was horrible. But Lovable knows their customers. They are developers, and developers are masochists anyway. They expect code not to run. They expect to get errors, and timeouts, and heavy load because that is their life. Now, obviously that is a very biased statement and I'm painting with a broad brush - but that's after 30 years of working with them. So give me a freaking break. I know these people, and they are insane. 🤣

If a person works on your house and screws up the paint job, and then says "I need to redo it. I need another $500" you would throat punch them three times before you could get ahold of yourself. Yet, if these vibe coding engines write code and it doesn't compile, it ask you "should I fix it"? And you're like "Of course! You wrote it, you own it". Then it charges you another credit? In the words of Joe Biden, "C'mon Man!"

Lovable has a pretty set template of what it likes to do, and what it won't do. If you are wanting to build a fairly straightfoward React based application with some Typescript, tailwind CSS, vite, and shadcn/ui graphs pointing at Supabase - you are in luck. But Lovable will not build lots of other cool stuff you might want to play around with, as I wanted to do.

I found one of the things Lovable does not like to do is deal with theme colors (like having both a light and dark theme). As you as you introduce that second theme, you will waste hours trying to get it to apply that theme to all pages, sub pages, and page objects. That to me would seem trivial to set colors and fonts, but it seemed a daunting task for it to comprehend. It also struggled with the navigation panels. At one point, it had three nested navigation panels on the left, and could not understand I wanted to have ONE panel, but different options to show based on user access permissions.

Now, having said all of that as a performance engineer, I need to switch hats to just someone who likes to build stuff. I spent close to 100K to try and build an application about 7 years ago that ultimately failed right before the finish line for an MVP to show customers. The final piece I needed was a dashboard with charts and graphs. You can read the whole story here. Even through all of the painful retries to get a request through under heavy load, I was able to create a reasonable web application that I could use as a starting blueprint to show developers for a REAL MVP. I would call it a starting prototype. And I did it in 3 hours. For Free. I am now thinking of reviving it as "SIMPLE RUM". I still own the domain for simpleapm.com and simplerum.io - so if anyone wants to help me build this out and make a bazilion from vibe coding - let's talk. But, I digress.

For that feat alone, it's worth the price alone. Purists may not like it, but this is the future. If I can get a simple application scaffolded without having to much more than wait on this thing to get my order right, then I would rather do that than bother with a developer until I have something to show them. These applications become developer versions of a slide deck pitch. I would never trust these applications, but if I can hand the IDEA over to someone who can really make it work - this is the best way I know to do it.

So to all who participated - congratulations, you were playing the role of a "virtual user" in a load test by Lovable, and they got your time for free on a weekend. Now they can go have a meeting and decide how to actually scale this application. They can also get a lot of ideas on what people want to build these days. Mining that data will be like a pot of gold for the marketing department. Contgratulations Lovable for creating a win/win.

If you want to find out about more stuff that I like to talk about, check out my website at scottmoore.consulting or my Youtube channels at youtube.com/@scottmooreconsultingllc or the new AI Testing World at aitestingworld.com and youtube.com/@aitesting world.

What's your take?