20VC: Base44's Maor Shlomo on How Vibe Coding Will Kill SaaS and Salesforce | Why it is BS that Vibe Coding Platforms Do Not Have Defensibility and Bad Margins | Why He Worries About Google, Not Replit and Lovable | Why Long Anthropic, Not OpenAI?

Maor explains his decision to sell Base44 to Wix for $80M rather than raise venture capital. He details how Wix provided the perfect combination of marketing muscle, infrastructure support, and operational resources while allowing the product team to stay lean and independent. The deal was structured with revenue-based milestones, and Base44 has since grown to over $100M in revenue.

• •Started Base44 as a bootstrap business after a capital-heavy previous company, grew to 100K+ users profitably
• •Chose Wix over raising VC because they target the same audience and have best-in-class marketing team
• •Structured deal to maintain product team independence while leveraging Wix's infrastructure, support, and legal/ops
• •Deal included financial compensation for hitting additional revenue targets and milestones
• •Base44 has grown to over $100M revenue post-acquisition, validating the decision

Maor makes the controversial argument that vibe coding will replace traditional SaaS tools like Salesforce and Monday.com. He believes it will become easier to build custom CRMs than buy licenses, especially for small businesses that don't need feature bloat. Software will become more 'liquid' with templates and customization replacing one-size-fits-all solutions.

• •Predicts vibe coding will become the largest software category, folding in CRM, support, task management
• •Small businesses prefer simpler, customized solutions over bloated enterprise features they don't use
• •Started Base44 to solve his wife's small business CRM needs - existing tools were 'hell on earth' to configure
• •Software requiring implementers (like Salesforce) won't survive in a decade
• •Smaller CRM companies and 'front-end on database' tools without moats will be eliminated first

Addressing criticism that vibe coding has no moat, Maor distinguishes between building a simple tool versus a platform for complex, production-ready applications. Base44's defensibility comes from vertical integration - building a 'mini cloud' with databases, user management, integrations, and scheduled tasks that handle millions of lines of code.

• •Easy to build a vibe coding tool for websites/front-ends, very hard to build platform for complex applications
• •Base44 applications reach millions of lines of code - requires heavy infrastructure and integration layers
• •Built vertically integrated platform with databases, user management, authentication, analytics, integrations
• •Used custom backend instead of third-party providers like Supabase - very hard for competitors to replicate
• •Moat is in handling complex projects that users actually use day-to-day, not just demo apps

Maor reveals he's not concerned about direct competitors like Replit, Lovable, or Bolt. His biggest worry is Google - if Gemini wins the model race, Google has the entire stack (compute, cloud, data, integrations) to dominate vibe coding. He emphasizes the importance of a competitive multi-model market for platform players like Base44.

• •Not worried about Replit, Lovable, or Bolt - they serve different segments (technical vs. non-technical)
• •Google is the real threat if Gemini wins - they have compute, cloud, data, and Google Suite integration
• •Market dynamics allow switching millions in spend overnight by changing one string of code
• •Uses different models for different tasks: Sonnet for design, GPT-5 for complex bugs, Gemini for certain workloads
• •Hopes for continued competitive race between model providers to maintain platform advantage

Maor addresses margin concerns in AI businesses, arguing margins are the least of his worries because model prices are trending toward zero. He discusses intelligent routing between expensive frontier models and cheaper open-source models, and how strategic decisions around optimization differ when you assume costs will drop dramatically.

• •Margins can change dramatically overnight based on product decisions and model selection
• •Strategic bet: model prices will go to zero, so don't over-optimize current costs
• •Plan to use intelligent routing - simple prompts to cheap models, complex ones to frontier models
• •Smaller open-source models getting better, will handle growing percentage of requests
• •Cursor's strategy: grow fast with expensive models, then introduce efficient proprietary model (Composer)

Maor explains how Base44 measures success differently than traditional SaaS. They track sentiment analysis of user messages in real-time, measuring whether the coding agent does what users ask. The focus has shifted from bug rates to whether agents make correct changes without breaking other features.

• •Measure sentiment of chat messages - users talk to AI like humans, expressing satisfaction or frustration
• •Track sentiment on tens of thousands of prompts per minute to gauge coding agent performance
• •Metrics evolved from bug tracking to measuring if agent does what user asks without side effects
• •Can see immediate impact when releasing upgrades or switching models (e.g., Sonnet 4 to 4.5)
• •Success is about users reaching actual day-to-day usage, not just publishing functioning apps

As an angel investor, Maor focuses on vertically integrated businesses that won't be commoditized by model providers. He advocates for investing in 'unsexy' industries like finance, restaurants, and healthcare where entrepreneurs build end-to-end solutions, rather than fast-growing but thin AI wrappers that lack moats.

• •Invest in vertically integrated businesses that build infrastructure, not just prompt engineering
• •Opportunities in 'unsexy' industries: law firms, hospitals, banks optimized with AI from the ground up
• •Avoid businesses where most value is in how they prompt - will be commoditized by model providers
• •Example: Building an AI-native law firm is more defensible than building legal AI software
• •Would invest in Cognition over Cursor at current valuations - more room to grow, less winner-take-all

Maor predicts 95-100% of code will be AI-written within two years, up from today's 50%. He argues that as models improve, even technical users won't want to see code - they'll prefer prompting in flow state. Base44 already sees ~90% AI-written code because the repository was structured from day one for LLM navigation.

• •Predicts 95-100% of code will be AI-written in two years (vs. 50% today at companies like Salesforce)
• •Base44 already ~90% AI-written because repository structured for LLM navigation from the start
• •As models get faster and better, even technical users won't want to manually edit code
• •Growing portion of software can be built without seeing code - users prefer prompting in flow state
• •Strategic decision: don't over-optimize what future model versions will solve automatically

Maor's advice for non-technical people building with AI tools: embrace iteration and don't be afraid to throw away code. The biggest mindset shift is understanding you're not working with a human - it's emotionally fine to revert and start over. Build for your own problems first, and iterate based on what the AI generates.

• •Build for your own problems first - easiest way to create actually usable software
• •Embrace throwing away code - it's emotionally hard but necessary for fast iteration
• •Start with high-level prompts, see what AI generates, then iterate on ideas you like
• •Don't plan like traditional development - save time by reverting bad features instantly
• •The ability to easily revert and start fresh is hard for people to grasp but crucial for success

Maor reflects on personal lessons from building Base44: money's value is in time, helping others, and taking bigger risks. He advises choosing partners similar to yourself, especially fellow entrepreneurs who understand the building mindset. He also shares the emotional difficulty of early platform instability and wishes he'd known those bumps would pass.

• •Money's value: buying back time, helping people you care about, and funding bigger future bets
• •Partner selection: choose someone identical to you who understands the entrepreneurial lifestyle
• •Early Base44 had infrastructure incidents that felt devastating but were just bumps in the road
• •Consumer products have wider emotional feedback spectrum than B2B SaaS - harder to handle
• •Regrets from previous company: missed twenties trying to compensate for inexperience with overwork