The Twenty Minute VCWix Founder: Will Base44 Win the Vibe-Coding Wars? | The Truth About the Economics of Vibe-Coding
CHAPTERS
- 0:00 – 4:12
Avishai’s nocturnal CEO routine and why it helps the org self-solve
Avishai shares his unusual sleep schedule (5–6am to ~11:30am) and explains how it creates uninterrupted thinking time. He also argues that arriving later forces teams to resolve smaller issues themselves, leaving only the hardest problems for the CEO.
- •Sleeps 5–6am, wakes ~11–11:30am; long-standing, likely genetic chronotype
- •Late hours create focused, meeting-free time for real work and planning
- •Teams learn to solve routine conflicts/issues without escalating to CEO
- •CEO time is reserved for higher-complexity decisions and people challenges
- 4:12 – 5:57
Why the market undervalues Wix: SaaS multiples, AI fear, and “trading on others’ news”
The conversation turns to Wix’s valuation versus its revenue base and the broader “SaaSpocalypse.” Avishai attributes much of the multiple compression to investor uncertainty about how AI changes SaaS defensibility, noting Wix often moves on OpenAI/Anthropic/Google narratives rather than company-specific execution.
- •Wix market cap vs revenue framed in context of SaaS multiple compression
- •Market struggles to price AI-driven risk to SaaS business models
- •Some SaaS names have rebounded; others remain down, fueling confusion
- •Wix stock often reacts to broader AI/company news rather than Wix-specific updates
- 5:57 – 6:57
Base44 gets “negative credit”: why the acquisition isn’t reflected in Wix’s valuation
Harry challenges why Base44’s fast-growing ARR seems to add little to Wix’s market cap. Avishai argues it’s worse than “zero value,” suggesting Base44 alone could command a much higher multiple than implied inside Wix, and that public markets cycle between being right and wrong.
- •Base44 ARR scale discussed as substantial within Wix’s overall story
- •Avishai claims Base44 would likely be valued far higher as a standalone asset
- •Public market logic can be cyclical; mispricings happen in both directions
- •AI uncertainty distorts valuation comparisons vs peers in “vibe-coding”
- 6:57 – 9:28
Why Salesforce and Atlassian are more resilient than people think (trust vs developer substitution)
Avishai differentiates SaaS defensibility: Salesforce’s moat is trust and data custody, not just CRM UI. He’s more cautious on Atlassian because developer customers are most capable of building replacements, even though market performance suggests there’s more resilience than he expected.
- •Salesforce moat: enterprise trust to host sensitive customer data
- •“You’re not going to vibe-code” institutional trust and compliance posture
- •Atlassian risk lens: developers are best positioned to build alternatives
- •Real-world stickiness suggests ecosystems/workflows are harder to displace than theory implies
- 9:28 – 13:13
Can SMBs really vibe-code their stack? Why Wix believes complexity still wins
The discussion focuses on whether AI reduces Wix’s TAM by enabling SMBs to build their own tools. Avishai argues most SMBs won’t (or can’t) build full business logic stacks; even Wix’s own teams struggled to prototype vertical workflows quickly, indicating meaningful complexity remains.
- •Counterargument addressed: AI could shrink TAM if many DIY their tools
- •Avishai: SMBs (pizza shop, hairdresser) won’t build full operational stacks
- •Internal test: building hairdresser business logic in Base44 proved slow/hard
- •View: future is segmented—some stay on Wix, some move to Base44, many hybrid
- 13:13 – 15:01
Base44 economics: margins, retention, and the push to improve cost with custom models
Harry presses on whether vibe-coding businesses have viable margins. Avishai states Wix’s core business is far more profitable and retains better, but Base44 costs are falling; Wix is prioritizing quality early while expecting cost to become less material as models and infrastructure improve.
- •Wix margins materially higher; website generation/hosting near-zero marginal cost
- •Base44 has higher inference costs; retention dynamics differ vs Wix core
- •Cost is declining via optimization and model strategy
- •Tradeoff: prioritize quality over cost in early market formation
- 15:01 – 17:03
Why Wix built its own AI model: domain data, better outcomes, and iterative training loops
Avishai explains the strategic case for custom models: Base44 has proprietary interaction data that can fine-tune performance for specific creation tasks. He contrasts generic frontier breadth (e.g., poetry knowledge) with the need for better task understanding in product workflows, and notes Wix already runs trained models for website generation.
- •Two drivers: improved quality for domain tasks + cost control
- •Base44 captures rich data on user intent, failures, and prompt mis-specification
- •Goal isn’t to replace frontier models; it’s to outperform them for Base44 jobs
- •Wix already uses trained models for website generation with weekly improvement loops
- 17:03 – 22:08
How much cheaper is “own model” really? And why quality still beats cost right now
They unpack the economics: Avishai cites modest savings (varying by task), challenging the common narrative of order-of-magnitude reductions. He argues the industry isn’t yet in pure cost-optimization mode—winning comes from better results and user experience, especially in a new category.
- •Cost savings vary; larger gains possible on simpler tasks and smaller models
- •Complex tasks like Base44 limit how far costs can drop without quality loss
- •Current priority: ship better capability rather than optimize for margin early
- •Base44 numbers disclosure discussed; Wix avoids over-segmenting metrics due to customer movement between platforms
- 22:08 – 25:40
M&A and the “one-person, $80M” Base44 deal: board logic, integration, and execution constraints
Avishai describes how the Base44 acquisition was approved with surprisingly little pushback on it being a one-person company. The board focused on strategy, go-to-market, team-building, and competitive differentiation; Avishai notes the bigger constraint now is operational execution, not financing, especially while launching new products.
- •Board diligence centered on business logic, GTM, and team scaling—not headcount optics
- •Base44 integration required building a company around the founder and adding Wix talent
- •Wix can fund deals with cash/FCF, but parallel integrations risk execution quality
- •Upcoming product launch shifts priority away from additional acquisitions
- 25:40 – 32:12
Capital allocation under pressure: buybacks vs acquisitions vs hiring (and the reality of timing)
Harry probes buybacks, take-private hypotheticals, and how public-market pricing affects strategy. Avishai frames buybacks as a valuable tool to counter dilution and effectively return capital, admits timing was poor, and outlines the classic triad of cash uses—acquisitions, buybacks/dividends, or hiring—each with tradeoffs the market reacts to.
- •Taking Wix private: Avishai declines to engage; focuses on building the company
- •Buyback rationale: excess cash, limited near-term M&A appetite, reduce float/dilution
- •Admits timing is hard; judges decisions on multi-year horizon, not quarters
- •Three uses of cash: M&A (hard, often fails), buybacks/dividends, hiring (hurts EBITDA)
- 32:12 – 34:24
Retaining talent when the stock is down: accept churn, protect top performers, refresh the org
Avishai argues there’s no magic retention playbook in a downturn; talent loss is inevitable and can be healthy if managed well. The key is maintaining a high concentration of top performers, using churn to refresh teams, and continuously developing the next generation of leaders and experts.
- •Downturns trigger inevitable attrition; even great companies lose people
- •Focus shifts from preventing departures to preserving “top talent density”
- •Churn can expose and elevate overlooked internal talent
- •Rebuilding pipelines matters: today’s stars once lacked experience too
- 34:24 – 38:20
Wix at 3,500 people: customer support reality, why off-the-shelf AI support fails, and AI skepticism
They discuss organizational scale: Wix employs ~3,500 people with customer support as the largest function due to global coverage needs. Avishai says they tried many external AI support tools without success and even internal attempts didn’t work well yet, cautioning that AI’s limitations are underestimated in real operational settings.
- •Headcount: ~3,500 total; Base44 ~400; support is the largest department
- •Global operations across 192 countries complicate “AI replaces support” narratives
- •Off-the-shelf AI support vendors underperform vs internal engineering-heavy realities
- •Avishai: AI is powerful but often fails in details; skepticism about near-term massive shrinkage
- 38:20 – 41:31
The real AI dream for SMBs: not replacing humans, but making small businesses dramatically more effective
Avishai reframes the upside: AI’s near-term impact is enabling SMBs to run better businesses—improving efficiency, sales, and customer outreach—rather than zeroing out headcount. He explains the “big but” of LLM trust: models can be unreliable in general, yet exceptional when constrained and tooled for specific workflows.
- •SMB upside: better operations, selling, and customer acquisition—soon, not in a decade
- •LLMs can hallucinate; users over-trust them, so domain tooling and guardrails matter
- •Examples of conversational AI as a newly normal capability (driving use case)
- •Differentiates use cases where AI works well vs where it’s the wrong tool (e.g., research)
- 41:31 – 47:54
Are SMBs excited or terrified by AI? Plus: career advice, law/health use cases, and limits of LLMs
Avishai observes SMBs are generally eager because they face daily operational pain and want help, while students may feel threatened by rapid change. The discussion broadens to where AI performs well (law, some diagnostics) versus poorly (scientific research without the right architectures), emphasizing that “LLM” isn’t synonymous with deep reasoning systems.
- •SMBs tend to welcome AI; they want relief from constant business problems
- •Students/university curricula lag: fear comes from uncertainty and slow institutional change
- •Law is a strong fit due to textual context and large corpora; medicine differs between diagnosis vs research
- •Frontier progress may require hybrid architectures beyond pure LLMs (e.g., simulations/specialized models)
- 47:54 – 59:07
Freedom, resilience, and staying present: handling storms, relationships, and parenting under pressure
The conversation shifts into personal operating principles: money provides freedom and the power of choosing to be in hard situations. Avishai shares resilience tactics (accept storms, focus on controllables, meditation/NLP), how he context-switches at home by time-bounding thinking, and a parenting principle of quality over quantity of time.
- •Money’s main benefit: freedom and the mindset of chosen commitment
- •Resilience: expect storms, do your best with what you control, reduce self-deception
- •Practical tools: meditation and NLP; mental reframing to reduce pressure
- •Relationships/parenting: time-bound work thinking; be present; prioritize quality time with kids
- 59:07 – 1:04:35
Quick-fire: changing views on AI, admiration for Figma, execution as the core risk, and Base44 scaling
In rapid-fire, Avishai says he’s become less convinced AI will replace humans as quickly as he feared due to persistent model shortcomings. He cites Figma as a rare product with near-religious user love, identifies execution as Wix’s biggest challenge, and estimates Base44 could grow to ~800–1,000 people depending on AI trajectory.
- •Changed mind: AI displacement likely slower than prior fears; AGI definitions shifting
- •Competitor admired: Figma’s product craft and intense user devotion
- •Biggest concern: execution—team efficiency and ambition against big goals
- •Base44 outlook: could double headcount with engineering-heavy composition; closes with gratitude and perspective
