What if I told you that advanced "thinking" AI models are just regular models that learned a new trick? In this session, we'll demonstrate the transformative power of fine-tuning.

We explore how chain-of-thought models—those that generate visible reasoning in `think` tags—are simply fine-tuned versions of standard transformers. This elegant example teaches us two critical concepts: how transformers use their own output tokens as input for subsequent predictions, and how fine-tuning can create entirely new model behaviors without architectural changes, and without requiring huge datasets.

Through live demonstrations, we'll discover:

* The fine-tuning workflow
* Why fine-tuned models outperform prompt engineering for specialized tasks
* The token-by-token mechanics that enable complex reasoning
* Common pitfalls and best practices from production deployments

Attendees will leave with practical skills and deep insights into why fine-tuning is the key to unlocking specialized AI capabilities.

***

Barry S. Stahl is a seasoned enterprise engineer and internationally recognized technical speaker with over three decades of experience in software development. He has delivered hundreds of sessions at premier conferences and user groups across North America, Europe, and Australia, including NDC Oslo, NDC Sydney, Craft Conference Budapest, Techorama Netherlands, and ConFoo Montreal. Barry’s engaging presentations span a wide range of topics—from microservices and software architecture to artificial intelligence and optimization algorithms—often featuring hands-on demos and real-world applications. His workshops and keynotes consistently draw enthusiastic audiences, and his reputation for clarity, depth, and practical insight has made him a sought-after speaker in both in-person and virtual formats. A curated selection of his past presentations is available on this [YouTube playlist](https://www.youtube.com/playlist?list=PLCo2TFzFXPTQ5qIZTbbzFNcJL348fl6uO), showcasing his dynamic delivery and technical expertise.