Do vibe coding tools (v0, Lovable, Replit) have any use for real engineers? Definitely - fast-iterating prototypes ALWAYS have a place in the dev cycle. Watch me try Claude Design while building slopwatch:

As always, the newsletter is the best spot for my up-to-date thoughts:

Matt, great as always!

best series so far.

Are these videos cross posted from yt or different content ? Thanks for sharing!

@nitrocitylab They will be, just seeing how they land first before deciding how I package them for YT

I tend to use these things when I don't have a clear picture on how a UI should look. Essentially the same way as you've done here. It helps generate ideas and can help lead thought. Real implementation is done completely separate after the picture is clear.

vibe coding prototypes is fine. shipping vibe coded prototypes to production is where it all falls apart

These tools are great for business to express what they want without having design skills and show to Dec team

Have you looked at stitch or

They are handly for the MVPs

I'd say they're handy to validate where you're going, not for producing production code. I.e. good for research

Alright, what's your go-to approach when you start using any of these? Personally, I never used repilt, but used v0, lovable and find it useful when starting with MVP and how it is gonna look alike in future.

The video above is a good example

Thanks, watching it out

Claude's design is quite compelling compared to v0. It could be the difference you were referring to: free vs paid service. I'm currently doing a frontend that is quite similar to what it generated and how you are generating it. I took an already built React theme and asked LLM to pick out the components as per the project's requirement written in the PRD.md file, but it feels like it's not working as of now. kimi-k2.5 btw

If 'slopwatch' is the target, then 'vibe coding' tools are clearly operating at peak efficiency.

The prototype value is obvious, but the bigger unlock is using it to explore API shapes and interaction patterns you'd normally skip because setup cost was too high.

looks promising. It looks like it basically allows you to prototype UIs using your design system with any coding agent or inside the Storybook itself

Some people keep saying AI is killing code quality. But looking at this, how is quality dropping when todays prototype looks like last years finished product? The baseline just moved

I still generate my landing pages first with vO…then pass it to Claude. Unfortunately I could not test Claude Design properly since I reached the weekly usage limit after less than one hour! 😮

yeah, v0's perfect for those rapid prototypes!!

Real engineers build their own in C + Linux and don't pay you're bullshit fees

When is the next stream for Slopwatch??

I’ve used it to spin up a working prototype for 2 projects I’m working on. I have no mobile experience so I had it start up the base prototype of a React Native app and now I can take that base and begin work on it. Replit made that possible for me. I don’t use it to build anything as a final product, but it’s incredible for quickly building out and seeing an idea come to life, and if it sticks, take it as the base prototype and then build on top of it. There’s a place for these tools.

You used like none of the cool features, I feel like you're not really this product's target audience

Which cool features?

Commenting on ui elements, asking for different knobs, handing off to claude code. There's probably more I don't know about 😅

Gotcha. Yeah I didn't want to bikeshed, I just wanted to get a sense if my language was correct and see an example of where we might be headed

While they do work well as demos/prototype before the actual implementation, i don't think for devs these tools have much value.

The real test isn't can you build it fast, it's can you maintain it six months later when you need to add that one feature users are screaming for. I've seen AI-generated code turn into spaghetti the second you go off its happy path. What's your move when Claude's structure doesn't match where the product needs to go?