Building Lovable: $10M ARR in 60 days with 15 people

Executive overview

Lovable is an AI software engineer that takes a text prompt and builds a fully working product in minutes. Users can then iterate and launch to the world. The product reached $10 million ARR in two months with just 15 people, demonstrating product-market fit. The core insight is that enabling non-technical people to build software removes the bottleneck that has prevented 99% of the population from turning ideas into reality.

Lovable removes the technical bottleneck for building software, enabling entrepreneurship at scale.

How Lovable works

• You describe an idea in plain English, and the AI generates a fully functional product
• The AI makes design decisions (like mimicking AirBnB) from minimal prompts
• Products are interactive and can be edited visually (like Squarespace), with code changes happening instantly
• Backend integration with Supabase allows adding login, data persistence, and payments
• The generated code can be deployed to cloud vendors like Vercel in one click

Scaling with AI: the technical breakthrough

• Most AI builders get stuck at certain points, unable to progress without human help
• Lovable identified the most critical bottlenecks (login, data persistence, Stripe payments) and hardened the AI to never fail on those
• The scaling law: invest in identifying where the AI gets stuck, then solve those problems systematically
• Users still need good prompting skills to unlock the full potential

The Lovable origin and vision

• Started as open-source tool called GPT Engineer after ChatGPT showed strong code generation ability
• Received 50K+ GitHub stars but shifted focus to building a product for non-coders
• Vision: build "the last piece of software ever"—a tool that makes all future software-building obsolete
• Immediate priorities: custom domains, team collaboration, helping founders reach users after launch

Growth and business metrics

• 300,000 monthly active users, 30,000 paying users, growing only through organic word-of-mouth
• Hit 4 million ARR in the first four weeks
• Reached 10 million ARR in two months with 15 people
• Fastest-growing startup in Europe; revenue now far exceeds 10 million ARR
• Made 1 million ARR per week for extended period after launch

How to be effective with Lovable

• Be patient and curious; use chat mode to understand how the tool works
• Be extremely specific about what you want—don't say "it doesn't work"; explain exactly what you expect vs. what you got
• Being a clear communicator and strong product manager is essential (arguably more important than technical skill)
• Spend a full week on an idea end-to-end to reach top 1% in AI tool proficiency
• Surrounding yourself with others obsessed about AI tools gets you to top 0.1%

Skills that will matter more and less going forward

• More important: discovering what to build, validating that users want it, having good taste and judgment about design
• More important: being a generalist who can learn multiple skill sets (architecture, design, product sense, user conversation)
• More important: understanding technical constraints and translating problems into technical solutions
• Less important: low-level implementation and writing code from scratch
• Still important: understanding software engineering fundamentals to make better decisions

Hiring and team building

• Lovable has 18 people (started at 15), with 12+ writing code at least part-time
• Look for people who care obsessively about the product, users, and team productivity
• Hire for superpower in one dimension (AI understanding, architecture, etc.) plus generalist learning ability
• High-ambition job postings attract the right people and filter out those seeking comfort
• Work trials of at least one day (sometimes a week) reveal how people actually think and contribute
• Office presence enables high-bandwidth communication and cross-pollination during lunch
• Raw cognitive capability is the strongest correlate for success at Lovable

Building product teams in the age of AI

• Everyone should be excited about using AI as a core part of their workflow
• The bottleneck is now taste and user understanding, not engineering
• Engineers should develop willingness to listen to users and understand what they care about
• Structure teams so non-engineer PMs don't have to understand every technical detail
• Weekly planning on most important problems keeps everyone aligned
• Use tools like FigJam and Linear to coordinate fast-moving teams
• Collaborate on obvious features (custom domains, team collaboration) while investing in longer-term initiatives (agentic behavior)

Lovable vs. competitors (Bolt, Replit, Cursor)

• Positioning: designed for non-technical users, not developers
• Key differentiator: visual editing (change text, colors instantly without code or waiting)
• Key differentiator: GitHub sync allows both Lovable and Cursor users to work on the same product seamlessly
• Reliability: most users report Lovable as the most reliable and least likely to get stuck
• Speed: other tools require asking the agent to make changes; Lovable lets you edit directly

What will change about software engineering

• Developers will shift to the top levels: understanding constraints, translating problems, maintaining quality
• The question "what should we build?" becomes more important than "how do we build it?"
• Better products will emerge because non-technical founders can now execute their vision
• An explosion of entrepreneurship as the technical barrier disappears
• Software quality will improve because builders can focus on solving real problems instead of fighting the tech stack

Lessons from past failure

• Worked at Thana Labs building an API to personalize learning (like Duolingo with AI)
• Product was technically good but failed because retrofitting AI into existing products is very hard
• Key lesson: start with the full user experience and ask where AI helps, not the other way around
• Adding cool tech to products doesn't work; solving real problems with AI does
• This applies to everyone building with AI: ask first what problem you're solving and for whom