The "open vs. closed" debate is a distraction. For the next generation of AI-native startups, the real moat comes with an agentic flywheel and owned expertise. While proprietary APIs offer a quick start, open-weight models and open-source tooling (from orchestration frameworks to local inference engines) let founders own their last mile of performance and UX. This shift recasts AI from a rented utility into a proprietary platform that backs a product experience.

In this roundtable, we dive into how founders are harnessing the open ecosystem to build agents that are faster, cheaper, and more deeply integrated than what’s possible via a generic API. In breakout groups, we’ll explore how local compute, open weights (like Nemotron and Qwen), and modular agent frameworks allow startups to compound value.

We’ll tackle the hard questions that founders face:
- Sovereignty: Why the "right to repair" your model is the ultimate defensive moat.
- The Latency War: How open stacks enable real-time agentic reasoning for industrial and enterprise use cases.
- Hybrid Realities: Orchestrating a stack that blends the brute force of closed models with the precision of fine-tuned open weights.