Stays from v3: Two-product structure. Asymmetric business model. 138 hotels in pilot, 405 tables, 501 pages production scale. Three-act framing of OS → Network → Standard. Mews $300M as category validator. Mandatory e-Arşiv as foreign-vendor exclusion. Türkiye AI Strategy funding alignment. €4.5M cumulative investment ask. 31 hires by month 22. 2,500 hotels by 2030. €350–600M ARR target.
Changes in v4: Brain architecture is now explicitly three-level — Industry, OC, Property — instead of just "property brain." Agent-native architecture is foundational not aspirational, with the MCP and agent surface pulled into Phase 1. Agent endpoint moat added as the seventh layer of defensibility. Model strategy is honest — Trendyol/Qwen/Llama/Gemma named explicitly, Turkish foundation model deployment positioned as add-when-ready not deploy-on-day-one. AI/ML budget revised up to ~€800K. Why-now expands to four forces with agentic-commerce inflection as the most strategically urgent.
The story is the same. The architecture is sharper. The pitch is harder to dismiss because it is more honest, more current, and more aligned with where the entire software industry is going.