20VC: Wix's Founder on What Wall St Gets Wrong About AI and Wix | Will Base44 Win the Vibe Coding Wars | The Truth About the Economics of Vibe-Coding | The Buyback Disaster: Lessons Learned with Avishai Abrahami

Summary

Avishai Abrahami makes the case that Wall Street is compressing SaaS multiples because it does not yet know how to price AI disruption, and he believes Wix’s core business is being undervalued in the process. He says Wix is still highly cash generative, with roughly $400 million in annual free cash flow, while Base44 is a fast-growing AI product with its own economics and a different cost structure. Abrahami is skeptical of the idea that vibe coding can replace complex business stacks like Shopify or enterprise systems like Salesforce, arguing that trust, workflow complexity, and product depth still matter. He also frames AI as most useful for SMB operations and support augmentation, not broad white-collar replacement, while acknowledging that future algorithmic jumps could change that materially.

Notes

• Wix is generating roughly $400 million a year in free cash flow and can fund Base44 internally.
• He says the market is pricing SaaS with AI fear, not just company-specific fundamentals.
• He argues the market assigns less than zero value to the core Wix business relative to peers.
• He says Salesforce’s moat is trust and data custody, not just the CRM interface.
• Wix’s customer base is mostly small businesses, which he says will not build full stacks with cloud code.
• Wix tested whether Base44 could build a hairdresser workflow and failed even with experienced developers.
• He sees the next five to six years as a hybrid world, with some customers staying on Wix and some moving to Base44.
• He says Wix is structurally more profitable than Base44 because generation and hosting costs are near zero.
• Base44 and Wix are building and fine-tuning their own models from product feedback and failure data.
• He says Wix already runs a trained website-generation model that is faster, cheaper, and makes fewer errors.
• He estimates their own model is between 1% and 30% cheaper than frontier models, depending on the task.
• In vibe coding, he says quality matters more than cost and the market is still early.
• He says Base44 does publish ARR occasionally, but not on a public-company cadence.
• Wix moves customers between Wix and Base44, and he says agencies are a key distribution channel.
• He frames the future as both products coexisting, not a replacement of Wix by vibe coding.
• He says he is not a stock trader and does not manage the business around market moves.
• Wix is public and he says its free cash flow is being redeployed into Base44.
• He admits the buyback timing was terrible, but says the rationale was excess float and cash.
• He still supports buybacks as a capital return tool and a way to offset dilution.
• He argues stock compensation is necessary to attract and retain talent.
• He says talent churn is inevitable, and the goal is high talent density rather than zero attrition.
• Base44 is around 400 people and Wix is about 3,500 people.
• Customer support is Wix’s biggest department, and he says global scale makes it impossible to run with 10 people.
• He says off-the-shelf customer support systems did not work, so Wix built its own.
• He says young AI startups can be outmatched by incumbent companies with deeper engineering and support.
• He says AI-only efforts at Wix repeatedly failed, so the company kept its core solution running in parallel.
• He says AI is powerful, but current systems are still overcredited for what they can do.
• He thinks SMBs will use AI to improve sales, operations, and profitability as soon as next year.
• He says LLMs are good at reframing and data collection, but weak at reasoning and prone to mistakes.
• He says legal work is a strong LLM use case because it is text-heavy, but medicine research is the wrong tool.
• He says K Health can diagnose better than most doctors, but diagnosis is not the same as research.
• He says astrophysics research needs simulations and multi-model systems, not a plain LLM.
• He says true scientific AI needs multiple models and architectures stitched together.
• He points to DeepMind’s protein folding as a real AI scientific achievement.
• He says money mainly provides freedom from fear and freedom to choose work.
• He says shocks are inevitable and reality changes in bursts rather than smooth trends.
• He recommends focusing on controllables and says meditation and NLP help him manage pressure.
• He says he is now less worried about immediate AI replacement than he was a year ago.