š Has anyone built a production app with bolt.new?
Iāve been playing around with it, and itās wild how fast it spins up a full-stack app from a prompt. But Iām wondering:
How well does it generate code for real-world use?
Any prompting tips to get better results?
How does it compare to others like Lovable?
Also, do you think tools like Bolt (full app generation) are the future, or will the Cursor-style coding assistant approach win out?
Would love to hear from anyone who's pushed this beyond a test project! š
Hey! We built the latest version of Tambor with bolt.new. It does a great job generating code that is almost production ready, but like all of these text to code tools, we find that it needs a good review, esp from a security standpoint, found some exposed API keys and such.
Compared to Lovable, we like the native integration with StackBlitz, because you can bring in your GitHub repos and it can work on those, so makes it easy to move it and out of bolt.new
I think these tools are 100% the future, and when you combine them with Agents, you get a real sense for how work moves. My prediction for dev related workflows in the next few years looks something like...Client or Stakeholder requests a tool, the request goes to an agent who scopes and builds 90% of it, then devs take a few days to secure it, check for bugs, agent does more QA/QC and in a week or two you do work that used to take 2-3 months.
That sounds awesome! š„ Totally agree that these tools are transforming dev workflows. The integration with StackBlitz and GitHub is a huge win for seamless iteration. And yeah, security checks are non-negotiableāexposed API keys are a real risk with these auto-gen solutions.
Your future prediction sounds spot on too. The combo of agents and devs refining work could drastically speed up delivery times while still keeping quality high. Exciting times ahead for dev teams! š
In my experience, I prefer cursor or windsurf to build applications, and even then, I usually stick to front-end development and prototyping to supplement my PRDs as a product manager. I am not sure if you can successfully build APIs that can be utilized in production using the above tools. Specifically, Bolt feels more focused on users with no technical skills and helps you build some basic website/web app designs, but Cursor and Windsurf allow you to make more sophisticated experiences, though they need you to be a bit familiar with programming concepts (just the very basics).
@hussein_rMostly, I build the V1 myself, incorporate feedback from stakeholders, and then have engineers handle integration (APIs, authentication, etc.) before shipping. This has cut back-and-forth with design and engineering by around 80%, as I can now prototype the ideal solution myself as a PM.
ā” is great if you want make something easy peacy,. But with $100 usage of o1 api-key in cursor you can build things that gets you funding from YC (ref todays launch https://www.producthunt.com/posts/lingodotdev) with 1 day of effort. Here is how: https://eoncodes.substack.com/p/how-i-built-an-ai-wrapper-saas-in
@sentry_co Thatās an interesting take! I think it really depends on what you're optimizing for, speed vs control. If you want something quick and functional, sure, but if you're aiming for something more robust and scalable, tools like Cursor with API access definitely open up way more possibilities.
Also, that YC-funded example is a great reference, crazy what can be built in just a day with the right approach!
Loving the excitement! š Iāve been messing around with it too, and the speed is insane. Definitely a game-changer for spinning up MVPs.
From what Iāve seen:
ā¢ Real-world code? It gets you like 80% there, but still needs manual tweaksāespecially for complex logic or scaling concerns.
ā¢ Prompting tips? Being super specific helps! Instead of āBuild a dashboard,ā try āBuild a Next.js dashboard with Tailwind, user auth, and a Stripe checkout.ā
ā¢ Compared to Lovable? Feels snappier, but Lovable seems more structured. Curious to hear what others think!
ā¢ Bolt vs. Cursor-style? Full app generation is wild, but I think devs will still want fine-tuned control. Maybe a hybrid approach wins?
Anyone taken this into a production-level project yet? Would love to hear some deep-dive experiences! š„
Tested both bolt and Lovable Bolt felt faster, but Lovable gave me more structured output.
I used Lovable to build the first version of my app landing page for DoubleMemory. Ejected after gotten the first version to Windsurf as it's easier and faster to iterate there. Never used Bolt but I assume it's similar.
I do believe some specialized envs will have it's days but it's too early to tell which paradigm will become dominant. The advantage of Bolt/Lovable is they can potentially integrate tightly with other service providers and have a network effect, it's harder for an editor to do that, although MCP may fix that in the future. With that said, their current form is quite limited, only few services were integrated with Lovable, maybe just supabase.
Havenāt tried Bolt or Lovable yet, but Iāve been using Cursor and itās already pretty impressive. Sure, you need to know your way around code, but the speed boost is crazy. Itās wild how much AI-assisted coding has improvedādefinitely curious to see how full app generation tools compare! š
From my personal experience, Iād say itās pretty handy for building the skeleton of an app, website, or design. But when you throw a ton of data at it, it starts to glitch out and burns through tokens way too fast. For me, itās more of a tool to cut down that initial time at the start of developmentāafter that, I just switch back to my usual workflow! š
You can also try this one https://github.com/stackblitz-labs/bolt.diy
I built grepjob.com with bolt.new. I'm super happy with it overall and I'm getting great feedback from users for the design.
My biggest gripe with bolt is it really needs 2 way sync with github. I want to be able to seamlessly switch between bolt and cursor. I also need git to really be productive. Bolt is a bit over eager and frequently overwrites files, so git would be a huge help to protect against unwanted changes
Product Hunt
Haha, you know I'm jumping into this convo.
I personally LOVE Bolt as a jumping off point. I think if you subscribed you can definitely get production ready apps (I've seen some on X). My thoughts on your questions:
I feel like the front end code is pretty solid and when connecting with @Supabase it works really well, however, the more complex the back-end, the less "real-world" ready I feel that it is.
I started working through my ideas with ChatGPT and then once it understands what I want, the steps to get there, and the tech stack I'd like to use I have it give me a prompt that I can "give to my AI Engineer" and usually helps kick things off in a good direction!
I actually use both in Tandem. I found @Lovable really good at front end and quick live-to production app but @bolt.new is nicer for developing a bit more complex apps.
I do think it's possible to develop production ready applications with both Bolt and Lovable and have seen folks ship! Even some launch on Product Hunt! How pretty the code is.... well that's something I cannot comment on. But he if it works... right?