CxC Ep37: Is Open Source the Future of AI? - with Ahmad Osman & Michael Bradley (Osmantic)
July 30, 2026 · 1 hr 13 min
Why listen
Two AI infrastructure founders explain why owning your AI stack beats renting it from frontier labs.
Enjoy the episode?
Subscribe to Code and Connor
In this episode
- Why open source models are closing the gap on closed source frontier labs
- How self-hosted AI cuts costs while keeping your data and IP in-house
- Why smaller specialized models can outperform large general models for enterprise work
- What Microsoft's shift toward open models signals about the future of enterprise AI
About this episode
This week on Code x Connor, we're joined by Ahmad Osman and Michael Bradley, co-founders of Osmantic, to discuss why open source AI is rapidly closing the gap on frontier models and how owning your own AI infrastructure could become a major competitive advantage.
Osmantic builds open source infrastructure that enables organizations to deploy, manage, and operate self-hosted AI systems, helping enterprises retain control of their data, reduce costs, and build specialized AI tailored to their own business.
In this episode, we discuss:
→ Anthropic’s accusations that Moonshot trained Kimi K3 on Claude
→ The rise of open source models and the threat it poses to closed source frontier labs
→ Why enterprises should own their AI infrastructure instead of renting intelligence
→ How self-hosted AI can reduce costs while protecting intellectual property
→ Why smaller, specialized models may outperform frontier models for enterprise workloads
→ The financial advantages of treating AI infrastructure as a capital investment
→ Why AI value comes from engineering, workflows, and context—not just larger models
→ How frictionless developer tooling could accelerate enterprise adoption of self-hosted AI
→ Microsoft's shift toward open models and what it signals about the future of enterprise AI
→ Why the next wave of AI innovation may come from open source rather than frontier labs
→ How infrastructure, hardware, and model efficiency could reshape the economics of AI
Be sure to follow this channel for insights and perspective from industry leaders, and schedule a meeting with us through www.codestrap.com to learn more about our offering.
Osmantic
Codestrap
Codestrap (X): https://x.com/Codestrap_ai
CodeStrap (LI): https://www.linkedin.com/company/codestrap
About Code and Connor
Code & Connor is where Connor Deeks and Dorian Smiley ask the questions most AI podcasts won't. Two founders — one strategist, one engineer — in honest conversations with executives, builders, and practitioners about what's actually happening at the intersection of AI and the enterprise.
See all episodes →
