This isn’t a post about whether you should use AI for coding or not. By now, it’s safe to assume that doing so is here to stay in some shape or form.
So, after experimenting with every AI-powered app builder we could find (Bolt, Loveable, et al.), our team was surprised by their popularity among developers, despite of their clear limitations. They are generally limited to building SPAs on top of Supabase. While that can make a lot of sense for basic apps, as developers we found these platforms quickly become limiting when you need to build anything with infrastructure beyond what Supabase offers, or use more complex architectures.
Another practical concern is that some of these tools don’t support proper isolated test environments, which significantly limits your control over deployment flows. For instance, approving a buggy SQL migration suggested by the LLM could inadvertently affect your production database.
These limitations aren’t necessarily flaws, as we suspect these tools might intentionally be aimed at non-developers who prefer simplicity and who may not be able to make use of more advanced features anyway.
At any rate, we wanted something different for ourselves, something made for us as developers.
So we set about creating a new tool, Leap, specifically for developers who want to make use of AI but need control over their architecture, APIs, infrastructure, and cloud deployment.
There’s a demo video showing Leap in action on the website: <a href="https://leap.new/" target="_blank" rel="noopener nofollow">leap.new</a>
So what makes Leap different? The workflow is similar, in that you start from a prompt, but the rest is pretty different:
<ul>
<li>You can iterate in a controlled way using versions and diffs. When connected to GitHub, approving a version will push a commit.</li>
<li>Apps are built using Encore.ts[1] for the backend implementation, it’s an open-source backend framework we created, already trusted by thousands of developers and with 9k stars on GitHub. The framework enables generating architecture diagrams and API documentation in real-time, so you can understand what you’re building even if most of the code is being generated using AI. (You can still make manual code edits of course.)</li>
<li>The framework provides a declarative infrastructure layer, sort of like a cloud-agnostic CDK, which means Leap is able to set up infrastructure for microservices, databases, pub/sub, etc., for each new change in ~1-2 seconds. This means you’re not iterating against your prod infrastructure at all, the preview environment is completely isolated.</li>
<li>For deployment, you can either take the code and use Encore’s open source tools to package your app into Docker containers, giving you the freedom to deploy anywhere. Optionally you can use Encore Cloud (this is our commercial product) to orchestrate deployments and infrastructure provisioning in your cloud on AWS/GCP.</li>
</ul>
We don’t intend for Leap to replace all current workflows and tools. For now, we expect it to be primarily useful for quickly setting up new projects or creating new systems in an isolated domain as part of an existing system.
We built Leap primarily because we felt existing tools didn’t match our needs as developers, but we’re just starting this journey and genuinely want to hear your thoughts.
<ul>
<li>Does this approach solve real infrastructure and deployment pain points you’ve experienced?</li>
<li>What else would you need to confidently use something like this to create production applications?</li>
</ul>
Your feedback will inform how we shape Leap, thanks in advance for taking the time to help us make something valuable for developers.
[1] <a href="https://github.com/encoredev/encore" target="_blank" rel="noopener nofollow">https://github.com/encoredev/encore</a>

The official daily.dev AI community. Led by thought leaders and AI experts. Stay tuned for insights, news and discussions. 

AI-powered app builders are popular but can be limited for developers, particularly in managing infrastructure and deployment. To address these issues, a new tool called Leap has been created. Using Encore.ts for backend implementation and providing a declarative infrastructure layer, Leap offers controlled iteration, isolated test environments, and flexible deployment options. Developers can still make manual code edits, and the platform supports GitHub integration for version control.

After trying every AI app builder that treats devs like no-coders, we’re building our own

Can’t wait for launch, Good Luck for the product.