Ahmad Osman, founder of Osmantic, argues that local AI is rapidly closing the gap with cloud-based frontier models. Speaking after running two workshops at AI Engineer World's Fair, he explains that running AI locally is no longer just about the model — it requires a full infrastructure stack including search, tools, and agent harnesses. He highlights that modern phones and laptops can now run models that outperformed cloud systems just two years ago, and that enterprises are increasingly motivated to control their own compute for data privacy, compliance, and resilience against provider policy changes. Osman also predicts that smaller, specialized models fine-tuned on company data will become the dominant enterprise pattern, and that open source model labs will monetize through licensing for fine-tuning and specialized deployments.
