Solo founder, $80M exit, 6 months: The Base44 bootstrapped startup success story | Maor Shlomo
Mayor Shlomo, a solo founder, shares his journey of building Base44, an AI app-building platform, and selling it to Wix for $80M in just six months. He discusses his tech stack, productivity as a solo founder with ADHD, and unique growth tactics.
Deep Dive Analysis
14 Topic Outline
Introduction to Maor Shlomo and Base44's rapid success
Origin story: Identifying the need for an AI app builder
Solo founding and bootstrapping: Challenges and advantages
Productivity hacks and AI tech stack for solo founders
The decision to sell Base44 vs. staying independent
Distribution and growth strategies in the age of AI
Initial growth tactics: From first users to viral spread
Building in public and community-driven growth
Impact of hackathons and strategic partnerships
Velocity as a key growth engine in product development
Technical stack and infrastructure insights for AI products
Counterintuitive lessons on user activation
The Base44 acquisition journey with Wix
Final advice for founders: Focus on passion and genius zone
5 Key Concepts
Batteries Included Approach
Base44's philosophy where every app built automatically includes essential features like a database, integrations, user management, and analytics, without needing third-party services or API keys. This simplifies the development of complex, functional applications.
Adaptive Software
Software built with AI that can be easily changed and adapted with natural language prompts as user processes or needs evolve. This allows for rapid iteration and customization without extensive manual coding.
Velocity as a Growth Engine
The idea that rapidly evolving a product and consistently releasing new features can itself drive user engagement and growth. Users get excited and attached to a product that is constantly improving, leading them to try it and share it.
Optimizing LLM Code Generation
A strategy to make Large Language Models (LLMs) write as little code as possible by providing a highly opinionated and structured code infrastructure. This reduces the chances of errors, confusion, and the need for LLMs to save extensive context for follow-up prompts.
LLM Routing for Code Generation
A pipeline approach where different LLMs are used for different tasks in code generation. For example, a powerful model like Claude 4 might handle initial prompt interpretation and UI design, while another like Gemini handles complex algorithms or debugging stuck loops.
10 Questions Answered
Base44 is an AI app building platform that allows users to describe what they want to build (an app, game, or website) using natural language, and AI codes it for them with a 'batteries included' approach, providing built-in database, integrations, user management, and analytics.
The idea stemmed from Maor's girlfriend needing a lead-capturing website and his experience volunteering with the Israeli scouts, where he saw a need for an easier way to build functional web applications without extensive coding or high agency costs, realizing LLMs could write the necessary code with the right infrastructure.
Benefits include potentially better financial outcomes if profitable, less stress without external funding pressures, and the ability to move faster with a small, focused team leveraging AI. Challenges include managing all operational aspects (DevOps, support), brutal prioritization, and the emotional burden of not having co-founders to share stress with.
Maor uses tools like RescueTime to block distractions, Cursor for coding, and custom AI-powered internal apps built on Base44 for tasks like content generation. He also optimizes his code repository to be LLM-suitable, allowing AI to write most of the code, and uses a mix of different LLMs for various tasks.
Maor started by begging three close friends to use the tool, sitting with them every other day to observe their usage, fix bugs, and build features for them. Once he saw them start sharing the product, he knew it was ready for broader exposure.
Key strategies included building in public on LinkedIn by sharing honest updates about the journey (good, bad, and ugly), leveraging a highly supportive community, incentivizing users to share what they built on social media with extra credits, and hosting a successful 'for good' hackathon that attracted sponsors and thousands of teams.
Base44's infrastructure was primarily built on Render.com for hosting and scaling, MongoDB for its flexible schema, and a Python backend. Maor controversially used plain JavaScript/JSX for the frontend instead of TypeScript, and kept frontend and backend in the same repository to provide better context for AI code generation.
Maor learned that getting users to an 'aha moment' as quickly as possible is crucial, even if it means sacrificing some initial 'helpful' features. He ditched a step where the LLM would first generate user flows before building the app because it slowed down the time to seeing the actual working application, making the surprise and conversion lower.
While Base44 was profitable and could have remained independent, Maor chose acquisition to 'play in the big league' and achieve global scale, believing that partnering with Wix (which shared similar DNA, customer base, and management chemistry) offered the best chance to lead the rapidly evolving AI app building category.
Wix reached out after community recommendations, and initial talks focused on advice before moving to acquisition. The due diligence was fast due to the company's newness. The deal was set to be signed on a Thursday night but was delayed by a few hours, leading to the signing occurring on Friday morning as a war between Iran and Israel broke out, adding unexpected stress to the final moments.
21 Actionable Insights
1. Build What You Love
Focus on building a product you genuinely enjoy working on and would use yourself. This makes it easier to work hard and stay energized over the long term, as it did for Mayor Shlomo with Base44.
2. Prioritize Deep Work
Optimize your workday to enable focused, deep work by using tools like Rescue Time to block distractions (e.g., social media). This is crucial for productivity, especially for solo founders with conditions like ADHD.
3. Automate Everything Possible
Invest time in automating processes, from code generation to content creation, to increase your pace and efficiency. As a solo founder, time is your most critical resource, so leverage AI and custom tools to do more with less.
4. Start with Close Friends as Users
For your first 3-10 users, recruit close friends and sit with them as they use your product. Observe their struggles, fix bugs immediately, and even build features for them to ensure early users find value.
5. Don’t Scale Before User Enjoyment
Avoid investing heavily in marketing until you see users genuinely enjoy your product and start sharing it organically. Sharing is the best metric for enjoyment and indicates you’ve reached early product-market fit.
6. Build Internal Productivity Tools
Leverage AI and no-code platforms (like Base44 itself) to build custom internal tools that automate your unique workflows, such as content generation for social media. This allows you to tailor tools precisely to your process and save time.
7. Focus on One Growth Channel
Identify a single marketing channel that shows promise and double down on it, rather than trying to be everywhere. For Mayor, building in public on LinkedIn proved highly effective, while other channels like paid ads did not work initially.
8. Be Honest Building in Public
When sharing your journey, be transparent about the good, the bad, and the ugly. This authenticity, combined with sharing learnings and progress (e.g., charts, numbers), can resonate deeply with an audience, especially fellow builders.
9. Incentivize User Sharing
Encourage users to share what they build with your product on social media by offering incentives, such as extra credits. This can create a powerful viral loop and expand your reach organically.
10. Velocity is a Growth Engine
Prioritize rapid product evolution and frequent feature releases. Users get excited and attached to products that are constantly improving, which can drive adoption and word-of-mouth growth.
11. Optimize LLM Code Generation
When building AI-powered coding tools, design a high-level, opinionated code infrastructure that minimizes the amount of code the LLM needs to write for new features. This reduces errors and improves efficiency by providing a clear framework.
12. Use Plain JavaScript/JSX for LLMs
Opt for plain JavaScript/JSX over TypeScript for front-end development when using LLMs to write code. Models find it easier to generate code in these formats, leading to faster and more accurate results.
13. Keep Front-End and Back-End in One Repo
Store both front-end and back-end code in the same repository when working with AI code generation. This provides the LLM with a more complete context, making it easier to understand and implement changes across the full stack.
14. Route Prompts to Specialized LLMs
Implement a pipeline that analyzes user prompts and routes them to different LLMs based on the task (e.g., Cloud for UI design, Gemini for complex algorithms). This leverages the strengths of various models for optimal results.
15. Accelerate to the ‘Aha Moment’
Streamline your user onboarding to get users to their core ‘aha moment’ as quickly as possible, ideally within a minute or two. Sometimes this means removing intermediate steps, even if they seem beneficial, to reduce friction and improve conversion.
16. Be a Person People Want to Work With
Cultivate strong interpersonal skills and ensure you are someone others would genuinely enjoy working with for years. This is crucial for partnerships and acquisitions, especially for small teams, as chemistry is a key factor for buyers.
17. Be Fine with the Alternative Path
When negotiating a deal (e.g., acquisition), be genuinely content with the outcome if the deal doesn’t go through. This mindset provides a stronger negotiating position and reduces undue pressure.
18. Use WhatsApp for Early Community Feedback
Leverage WhatsApp groups for early-stage community engagement to gather quick feedback and detect bugs. Its real-time nature makes it effective for immediate insights, though it may not scale for larger communities.
19. Host ‘For Good’ Hackathons
Organize hackathons focused on building applications for social good. This can attract a large number of participants, generate positive publicity, foster community, and even attract sponsors, serving as a powerful growth engine.
20. Choose Flexible Database for AI Apps
When building AI-powered applications, consider using a flexible database like MongoDB. LLMs may frequently change data schemas based on user requests, and a non-relational database can adapt more easily to these evolving structures.
21. Utilize Render.com for Infrastructure
For solo founders or small teams, use platforms like Render.com for infrastructure management. It simplifies deploying and scaling web apps, databases, and other services, reducing the need for extensive DevOps knowledge.
6 Key Quotes
I don't think I've written a single line of HTML or JavaScript in the past three months.
Maor Shlomo
I'm not going to try and scale anything before I know that users enjoy it. And the best metric to seeing them enjoying it is that they're starting to share it with someone.
Maor Shlomo
If you're able to show up every day, you know, your chances are like goes up immediately.
Maor Shlomo
The best position to negotiate such a deal or even to get there is to be also very fine with the other path of not getting acquired.
Maor Shlomo
Holy shit. Like it actually understood me. And you see the app. And if you have like a stage in the middle, it makes a surprise, like slightly less surprising.
Maor Shlomo
Just make sure that at least 50% of your time you work on some, on the parts of you that you really like and that you're really good at.
Maor Shlomo
3 Protocols
Maor Shlomo's AI-Powered Content Generation Process
Maor Shlomo- Write down high-level content ideas for the week.
- Input ideas into a custom Base44 app (or similar AI tool).
- The app, using a saved tone of voice and previous post context, generates a LinkedIn post.
- Review and approve the LinkedIn post.
- The app then breaks down and adjusts the content for a Twitter post/thread.
- Approve the Twitter content.
- Generate an image for the post using a separate tool (this step was done manually before the custom app fully integrated it).
Maor Shlomo's Early User Acquisition and Feedback Loop
Maor Shlomo- Identify 3-10 close friends or individuals who owe you a favor or have a strong reason to use the product, especially if they are currently unemployed and looking to build something.
- Get them to sit down with you every other day, physically observing them use the tool.
- As they try to build something and encounter issues, immediately review logs, make changes, and push updates to production.
- Build features directly for them based on their immediate needs and struggles.
- Do not invest in marketing until users organically start sharing the product with others, indicating genuine enjoyment and value.
Maor Shlomo's LLM Optimization for Code Generation
Maor Shlomo- Build a very high-level, opinionated code infrastructure that handles common functionalities like CRUD operations, authentication, and database interactions.
- When asking the LLM to implement a new feature, prompt it to write as little code as possible, leveraging the pre-built infrastructure.
- Use plain JavaScript/JSX for the frontend instead of TypeScript, as it is generally easier for LLMs to write.
- Keep frontend and backend code in the same repository to provide the AI with comprehensive context.
- Implement an LLM routing pipeline: use powerful models (e.g., Claude 4) for initial prompt interpretation and UI design, and smaller, faster models (e.g., Gemini, Flash) for complex problem-solving, algorithm generation, or patching code within files.