The Twenty Minute VCBase44’s Founder, Maor Shlomo on How Vibe Coding Will Kill SaaS
CHAPTERS
- 0:00 – 4:57
Why Maor sold Base44 to Wix: bootstrap success meets scale constraints
Maor explains Base44 started as a “for fun” bootstrapped return to coding, then rapidly hit massive adoption and profitability. He outlines the crossroads: stay solo and risk being outgunned by heavily funded rivals, raise capital, or join a scaled parent with aligned distribution and operations.
- •Base44 began as a bootstrapped side project driven by curiosity about LLM-driven software
- •Traction surged from tens of thousands to hundreds of thousands of users quickly
- •Three options: remain solo, raise aggressively, or partner via acquisition
- •Wix fit: same audience, strong marketing engine, and ability to keep product team lean
- •Deal structure designed to preserve upside via milestones and revenue-linked compensation
- 4:57 – 8:43
“Vibe coding will fold SaaS categories”: build-your-own tools vs buying licenses
Maor argues many classic SaaS categories (CRM, support, project management) will increasingly collapse into a vibe-coding/app-building layer. He believes software will become more customizable and “liquid,” making it easier in many cases to create tailored internal tools than to buy bloated one-size-fits-all SaaS.
- •Vibe coding as an umbrella category that absorbs CRM/support/project management over time
- •Future may be templates + iterative customization, not one prompt = full Salesforce
- •Ownership benefits: less lock-in, more control over workflows and data exposure
- •Feature bloat in traditional SaaS creates room for lean, bespoke alternatives
- •Enterprise shift will be slower, but SMBs may adopt faster due to simplicity needs
- 8:43 – 11:56
SMB reality check: tattoo-artist CRM pain as Base44’s origin story
Harry challenges whether small businesses care about owning code/data; Maor reframes the need as simplicity and fit-to-purpose. He shares how helping his then-girlfriend (now wife) build a CRM using existing “customizable” tools was painful, motivating Base44’s integrated approach.
- •SMBs don’t necessarily want code ownership—they want simpler, tailored workflows
- •Drag-and-drop configurable tools still feel like “hell” even for software-savvy users
- •LLMs can generate a leaner system if the platform provides the missing scaffolding
- •Base44’s thesis: remove setup friction (DB, integrations, email, etc.)
- •Traditional pre-2022 SaaS becomes harder to justify as customization gets easier
- 11:56 – 16:37
Defensibility in vibe coding: the moat is infrastructure, not demos
Maor distinguishes between flashy “clone in one prompt” demos and production-grade app building. He argues defensibility comes from deep platform layers—integrations, databases, auth, scheduled jobs—essentially building a mini-cloud that enables real applications at scale.
- •Easy to create a vibe-coding toy; hard to ship a platform for real-world complexity
- •Production apps require many prompts, deep context handling, and robust agent behavior
- •Moat = integrations + compute + platform primitives (DB/auth/user mgmt/analytics)
- •Base44 targets complex functional apps rather than simple websites/landing pages
- •Sustained value comes from day-to-day usage, not one-off project generation
- 16:37 – 23:18
3–5 year market map: incumbents add vibe layers; pro tools vs no-code builders converge
They explore how incumbents (monday.com, Microsoft, potentially Salesforce) will embed vibe coding inside existing systems of record. Maor predicts segmentation by user type (developers want code; non-technical users don’t), with gradual convergence as each side moves toward the other.
- •Incumbents will add in-product vibe tooling to increase extensibility and customization
- •Systems of record (e.g., Salesforce) may become configurable UI shells over core data
- •Base44’s goal: users never need to see a line of code; Cursor serves developers well
- •Capability frontier: some software (e.g., native firewall) remains hard for app builders
- •Roadmap expands “percent of software you can build” via primitives like scheduled tasks and bots
- 23:18 – 30:47
Who Maor fears: model-provider dominance and Google’s full-stack threat
Pressed on competitors, Maor says he worries less about specific builders and more about model-market dynamics. If one model provider wins decisively, it could logically move up-stack to dominate vibe coding; he singles out Google as most capable due to compute, cloud, and product integrations.
- •Competitive risk shifts from peers to upstream model providers if consolidation occurs
- •Multi-model strategies offer advantage—unless one provider wins by a wide margin
- •Google named as the most likely to move up-stack if Gemini dominates
- •Switching LLM spend is extremely fluid compared to cloud vendor switching
- •Race dynamics create rapid shifts in cost/performance and product strategy
- 30:47 – 33:54
LLM economics and switching: routing, open-source models, and why “margins don’t matter” (yet)
Harry pushes hard on gross margins and LLM COGS; Maor argues margins are highly decision-dependent and likely to improve. He describes intelligent routing (cheap model for simple prompts, frontier model for hard tasks) and expects declining model prices and stronger open-source options to drive better unit economics.
- •Exact LLM cost share not disclosed; economics vary with product design choices
- •Model routing/proxy: match prompt complexity to appropriate model for cost + speed
- •Open-source models improving may take a growing share of requests over time
- •Autonomous agents that run long jobs can burn cash if user intent is unclear
- •Maor’s stance: focus on growth and product value; margin optimization will follow as costs fall
- 33:54 – 36:20
Competing in weeks, not years: building features that are hard to copy
They discuss how AI compresses copy cycles, forcing a new product strategy. Maor says the answer is to place big bets on foundational platform components (like a built-in backend) while accepting that smaller UX wins will be rapidly replicated.
- •Feature copy cycles have shrunk to weeks; velocity alone isn’t enough
- •Base44’s major bet: vertically integrated backend rather than relying on Supabase/Neon
- •Owning core infrastructure is hard to replicate and hard for competitors to migrate toward
- •Other differentiation levers: UX taste, delight, and shipping ahead of the market
- •Example: “app suggestions” shipped and copied in days—acceptable in this new game
- 36:20 – 41:52
Defining user success: from bug rates to “sentiment in the chat”
Maor explains that “finished app” is an inadequate measure because many users can publish functioning apps; the harder question is sustained usage. Base44 uses a novel success metric—sentiment analysis of user chat prompts—to quantify whether the agent is doing what users asked and improving with model upgrades.
- •“Deployed and works” is saturated; day-to-day usage is the real bar
- •Tracks chat sentiment because users talk to the agent like a human (praise vs curses)
- •Measures negativity minute-by-minute across massive prompt volume
- •Model upgrades (e.g., Sonnet versions) show observable sentiment improvements
- •Platform primitives (e.g., scheduled tasks) expected to lift success metrics by removing blockers
- 41:52 – 50:59
Do early AI revenues matter? What Maor looks for as an investor
Harry asks how to interpret today’s extraordinary AI revenue numbers; Maor says the bar has shifted, but revenue alone isn’t enough. He prioritizes whether a company will be “eaten” by model providers and whether it can become vertically integrated, pointing to Cursor’s move into proprietary models as a strategic template.
- •Revenue growth expectations have reset (e.g., “zero to 10” ARR can be plausible and still impressive)
- •Key diligence: defensibility vs model-provider commoditization
- •Vertical integration as a hedge—own more of the stack, not just prompts
- •Cursor example: grow via frontier models, then improve margins with an in-house model
- •Jasper cited as a cautionary tale where moat was thin beyond LLM usage
- 50:59 – 53:01
Where “smart money” should go: unsexy vertical plays vs fragile agent wrappers
Maor argues the best opportunities are in less glamorous industries where founders can build end-to-end, AI-native operations. He’s wary of investing in agent companies if their advantage is mostly tool-wrapping that model providers can replicate inside ChatGPT/Gemini interfaces.
- •Underinvested: “not sexy” sectors (finance, restaurants, operations-heavy businesses)
- •Thesis: build full-stack, vertically integrated companies rather than thin software layers
- •Skepticism on generic agent startups—risk of commoditization by frontier labs
- •Invest only where you understand the domain well enough to judge moats
- •AI unlocks previously “bad ideas” (new law firms, hospitals, banks) as viable ventures
- 53:01 – 57:47
How Base44 beats Cursor: non-technical-first UX and invisible infrastructure
Prompted to make the competitive case, Maor positions Base44 as the place to build functional software without dealing with API keys, servers, or file trees. He claims that as models improve and iteration tightens, even technical users will increasingly prefer a fully integrated environment over managing code-centric workflows.
- •Base44’s wedge: enable serious apps without code visibility or developer setup steps
- •Early market friction: other tools required Supabase keys and technical configuration
- •Over time, fewer bugs + better models reduce the need for manual code editing
- •Thesis: “if you can build it in Base44, you’ll prefer Base44” due to speed and simplicity
- •The addressable portion of software buildable in Base44 expands as primitives and integrations grow
- 57:47 – 1:01:13
Will the AI boom slow? Why Maor expects continued value creation
They discuss fears of an AI bubble and revenue “speed bump.” Maor believes the ecosystem is still early in extracting economic value—even with today’s models—and points to dramatically increased company-building efficiency as the driver of continued growth and downstream consumer benefit.
- •Key question: platform gains that lift the ecosystem vs value captured only by model labs
- •Maor doesn’t expect a speed bump; says we’re “scratching the surface” of value
- •LLMs enable tiny teams to build and monetize meaningful businesses
- •Efficiency gains should make products cheaper and better across the economy
- •Contrasts capital-heavy prior company experience with lean AI-native company building
- 1:01:13 – 1:05:07
Valuations, Wix vs startups, and operating outside Silicon Valley
Harry challenges the valuation gap between public Wix and richly valued private AI builders; Maor argues Wix is undervalued and emphasizes the moat of scale, marketing, and execution. He also says being outside Silicon Valley can help, given talent dynamics and Wix’s backing as a “Silicon Valley equivalent.”
- •View: Wix undervalued; scale and marketing capability are strategic moats
- •He doesn’t over-index on competitor noise; focuses on product + metrics improvements
- •Market is large enough that multiple players can win without winner-take-all dynamics
- •Not being in SV can reduce talent wars and distractions; world is increasingly distributed
- •Wix provides deep-pocket support and operational leverage without SV location requirements
- 1:05:07 – 1:18:29
Quick-fire: AI-written code, life after the exit, and founder lessons
In rapid-fire, Maor predicts AI will write nearly all code within two years and gives practical advice for non-technical builders: iterate fast, expect to throw away versions, and build for your own pain. He also reflects on money’s role, personal regrets from overworking, partner selection, and what he’d tell himself during early reliability crises.
- •Prediction: AI-written code rises to ~95–100% for many teams; agent prompting becomes the workflow
- •Advice: build for your own problem; iterate rapidly; revert/throw away early versions easily
- •Money helps mainly by buying time, enabling generosity, and funding bigger future bets
- •Personal reflection: regrets about sacrificing time in his 20s; overwork as inexperience compensation
- •Hard-earned lesson: early incidents and emotional feedback are “bumps in the road”—don’t internalize them
