Two creators from the company Lovable introduce and demonstrate the new, improved version of their AI agent. One of the creators, Caroline, uses the agent to build a Swedish language learning app from a simple text prompt, showcasing the AI's intelligent planning, automated testing, and prompt queuing features before publishing the final app.
Post caption
Lovable is now 71% better at solving complex tasks - building more on its own with less hand-holding.
Here’s what‘s new:
Smarter planning - Maps out features before coding, asks better questions, and shows you the plan every step of the way
Auto testing - Opens its own browser, tests like a real user, finds bugs, and fixes them on the spot
Prompt queueing - Stack multiple ideas while it works. No more waiting for your turn
1-click Google auth - Add sign-in instantly, no dev console setup needed
#lovable #tech #webdev #aitools #ai
Transcript
Hi. I'm Fabian from Lovell. And I'm Caroline. And today, we're introducing the next generation of Lovell with a bunch of new capabilities. And to show how it works, we'll walk you through what Lovable does as Caroline builds an app. No, Caroline, what are you going to build? Well, Lovable is a Swedish company based here in Stockholm, but I'm not Swedish like you are, Fabian. Mhmm. So I'm going to ask Lovable to build an app to help me learn the language. I'll start by telling Lovable to create an app to help me learn Swedish. By starting your project in plan mode, Lovable asks thoughtful questions to understand exactly what you need from features to user experience. Then it creates a detailed step by step plan for tier approval. And then when Lovable is finished with the first version, Lovable now best your apps like a real user. It can open a browser, navigate flows, investigate edge cases, and catch bugs. If it finds issues, it doesn't just flag them. It corrects them on the spot. Alright. So I just queued up three more tasks to add Google login, an AI voice chat, and a cool AI generated background video to my app. And to build that, love all yes situations. Our new Google auth for login, 11 labs for voice, and our own built in AI for generating the video. It seems like everything is working. Perfect. So now we can just hit publish, and then the one we've able to use. Almost. That's pretty good. Thanks.
