While loops with tool calls - Practical AI Recap

Podcast: Practical AI

Published: 2025-10-30

Duration: 45 min

Guests: Jared Zonnerich

Summary

The episode delves into the evolution of AI applications from complex workflows to simpler, more autonomous systems using while loops with tool calls. It highlights the importance of context engineering and how AI models have improved in handling tasks autonomously.

What Happened

The discussion begins with Daniel Wightnack and Chris Benson talking about the festive season and Halloween preparations, setting a lighthearted tone for the episode. Jared Zonnerich, CEO of PromptLayer, joins the conversation, reminiscing about the origins of the company's cake logo and discussing their plans to bring real cakes to conferences, illustrating a creative marketing tactic.

Jared highlights the evolution in AI prompting, mentioning the rise of reasoning models that have simplified the process by incorporating chain of thought into the models themselves. This transition has made AI systems more reliable and easier to guide, yet prompting remains central to leveraging AI effectively.

The conversation shifts to how prompting has evolved into context engineering, where the focus is on the quality and quantity of information fed to AI models. Jared emphasizes that AI applications now have more tools at their disposal, requiring careful consideration of context to prevent models from getting distracted.

The hosts explore the transition from complex directed acyclic graphs (DAGs) to more autonomous systems using while loops with tool calls. This new approach allows for greater flexibility and efficiency in AI applications, as models now have built-in capabilities to handle tasks like tool calling more effectively.

Jared discusses the challenges of keeping up with rapid advancements in AI, suggesting that users need to experiment with different models and approaches to determine what works best for their use cases. Testing AI applications involves a blend of heuristic evaluations and specific tests to ensure reliability and performance.

The episode delves into the importance of collaborating between technical and non-technical teams to successfully implement AI solutions. Jared points out that understanding the nuances of what makes an AI application good is crucial, particularly when involving domain experts in the development process.

The conversation wraps up with excitement over the future of AI, particularly the democratization of AI engineering for non-technical users. Jared envisions a future where AI tools are seamlessly integrated into workflows, enhancing productivity and enabling more people to build AI-driven applications.

Key Insights

• Reasoning models in AI have evolved to include chain of thought processes, making them more reliable and easier to guide while still requiring effective prompting to maximize their potential.
• The shift from complex directed acyclic graphs (DAGs) to systems using while loops with tool calls allows AI models to handle tasks like tool calling with greater flexibility and efficiency.
• AI applications now focus on context engineering, which involves managing the quality and quantity of information provided to AI models to prevent distractions and improve performance.
• The democratization of AI engineering is anticipated to enable non-technical users to integrate AI tools seamlessly into workflows, enhancing productivity and allowing more people to develop AI-driven applications.