With Model OS Founder Fernando Jia: The Doom of the “Rent Model” Era

From an investor's seat at Y Combinator China (Miracle Plus), Fernando watched AI startups get absorbed by the very models they were built on. Now he's building Model OS so they can own their intelligence instead of renting it.

On Episode 9 of 18VC podcast, we sit down with Fernando, founder & CEO of Model OS (Intelligence Cubed) — the "survival infrastructure" that turns AI apps from prompt wrappers into companies that own their own models, evals, routing, and feedback loop.

Highlights & Key Insights:

• Why he calls most AI apps "prompt wrappers" — and the 3 forces quietly driving a wrapper "extinction event"
• The exact day Model OS was born: Jan 20, 2025, when DeepSeek trained a frontier-level model for $6M while OpenAI spent tens of billions
• "Turnitin on chain" — how watermarking model weights pays the original creator a cut every time someone fine-tunes their model
• When a founder actually needs fine-tuning vs. just RAG — and the foundation-model mistake almost everyone makes first
• The 4 revenue layers behind a free-to-use router with 2.5M+ monthly active users
• His contrarian advice to young founders itching to drop out (graduate faster instead)