newsroom
Enterprise AI Adoption · Trust and compliance moats protect enterprise SaaS; SMBs want AI to run business, not build software
now playing · Enterprise AI Adoption
Wix builds own fine-tuned models at 1-30% cost of frontier models with better task-specific quality
Wix and Base44 train proprietary models on their own usage data (prompts, failures, successes). These narrow models outperform generic frontier models for website generation and app buildin…
Vibe coding overhyped for SMBs; Base44 targets developers/agencies, not pizza shops
Avishai tested building a hairdresser management system in Base44 with professional developers — failed after two weeks. Most SMBs lack technical skill to vibe-code complex business logic;…
Trust and compliance moats protect enterprise SaaS; SMBs want AI to run business, not build software
Enterprises buy trust (Salesforce, Palo Alto) not features. SMBs have too many operational problems to spend time vibe-coding; they want AI to automate marketing, scheduling, sales — not re…
LLMs hallucinate in medical/legal research; specialized architectures needed for high-stakes reasoning
General LLMs treat Reddit posts and Nature papers equivalently — dangerous for medical research. Legal works better due to textual nature. True scientific reasoning requires connecting LLMs…
SaaS multiples compressed because market cannot price AI disruption risk
Public SaaS stocks trade at depressed multiples not due to fundamentals but because investors cannot quantify which companies AI will disrupt. Trust-moat companies (Salesforce) are safe; de…
AI coding agents not yet cost-optimizing; quality trumps margin in early market
Base44 margins far below Wix (hosting ~zero cost vs GPU inference). Avishai prioritizes quality over cost — 20% better quality beats 20% lower cost. Own models cut inference cost to 1-30% o…