Transcript note: two short garbled stretches (a repetition loop on the open-source discussion, and a mixed-language passage near the end).
roy-lazar (Wonderful) explains how the company went from founding in January 2025 to a ~50–100M ARR within about 1.5 years, by being enterprises’ “primary applied-AI partner”. The thesis: AI delivers ~10× SaaS value because it does tasks rather than assisting people. Adoption lags outside the US, an arbitrage to capture. AI-level contract values (millions per year) make a local enterprise motion viable in each of 30 countries (e.g. Greece: ~25 enterprises × ~$5M). It started with voice/chat customer support (e.g. a Maccabi health-fund agent) and expanded into back-office work (mortgage processing at a bank, next-best-offer at an Italian telco, warehouse orchestration). Forward-deployed engineers install a product, not consulting hours. The aim is a horizontal “one AI” per enterprise (an AI OS per vertical), not point solutions. Internally: their own stack built ahead of open-source libraries, model-agnostic components re-benchmarked continuously, no hand-written code for six months, an internal “Wonder” agent answering any factual question, features built “half-broken” betting on model progress, and a regular “what would I do with 10× agency?” check.
Show: ignore-instructions
Concepts: enterprise-ai-adoption · second-brain · model-selection · voice-agents · ai-hype · ai-rd-rollout