At the NDC London 2025 conference, Blazor creator Steve Sanderson showcased the exciting future of .NET app development, focusing on smaller AI models and autonomous agents. He emphasized the potential of agentic AI, which allows systems to perform complex tasks autonomously, making it particularly advantageous for developers in complex environments. Sanderson also pointed out that smaller AI models can often be more effective than larger ones for many applications, enhancing efficiency and data privacy. A highlight was a demonstration of a Blazor voice assistant capable of understanding user intent and managing tasks seamlessly within applications. Sanderson’s insights pave the way for innovative user experiences and tools for developers working with AI in .NET.
Steve Sanderson Previews AI App Dev: Small Models, Agents and a Blazor Voice Assistant
In a recent keynote at the NDC London 2025 conference, Steve Sanderson, the creator of Blazor, shared insights on the future of .NET application development. He focused on the integration of smaller AI models and autonomous agents. A highlight of his presentation was a new Blazor voice assistant, which showcased exciting new capabilities.
Sanderson’s keynote titled “The past, present, and future of AI for application developers,” emphasizes the shift towards agentic AI. This new approach allows systems to perform multiple tasks autonomously, moving beyond simple single-prompt interactions.
Understanding Agentic AI
Sanderson explained that agentic AI systems can reason, gather context, and evolve towards solutions by handling multiple tasks independently. This change is particularly significant for developers involved in creating complex tools or automating DevOps processes. For instance, a software upgrade agent could research new updates, analyze code, and suggest modifications—all on its own.
“This is an area many developers are actively exploring,” Sanderson noted, reflecting on the growing trend towards agentic systems.
Small Models Matter
Sanderson urged developers to consider using smaller language models instead of massive ones like GPT-4 for many applications. Compact models can handle tasks like intent classification or data extraction effectively. Their lightweight design means they can run on local servers or even directly in the user’s browser, enhancing performance and ensuring data privacy—especially crucial for internal business applications.
He demonstrated this concept using a real-world example of a lightweight classification model that operates seamlessly in a browser, showing that effective small model solutions are not just theoretical.
The Blazor Voice Assistant Demo
One of the standout demonstrations was a real-time voice assistant created with Blazor Server and OpenAI’s API. This assistant not only recognized speech but also understood user intent, managed UI data, and called backend functions. The demo illustrated how smoothly it integrated into the .NET framework, offering an exceptional user experience that devs can explore today.
“This assistant is part of the application workflow, providing valuable support to users,” Sanderson explained.
Key Takeaways from Sanderson’s Keynote
Audiences could learn more about:
- Creative AI from its roots with Alan Turing to advanced language models.
- Training a mini GPT-2-style model live using session data.
- How to leverage AI to call external functions through the right prompts.
- A document-based AI assistant that queries local PDFs with citations.
- Structured data extraction in .NET applications.
- AI tools for image analysis and traffic management.
In summary, Sanderson’s presentation highlighted a transition towards smaller, more efficient AI models and real-world applications that are ready for developers today. As AI technology evolves, there is a great opportunity for .NET developers to create innovative solutions that shape the future of application development.
About the Author
David Ramel is an editor and writer at Converge 360.
Tags: AI application development, Steve Sanderson, Blazor, autonomous agents, small language models, voice assistant, .NET solutions.
What are Small Models and Agents?
Small models and agents are simplified versions of AI programs. They perform specific tasks like answering questions or recommending products. They are designed to be lightweight, making them easy to use and efficient.
How do Small Models improve the Blazor Voice Assistant?
Small models make the Blazor Voice Assistant faster and more responsive. They help the assistant understand voice commands better, allowing it to listen and respond quickly. This leads to a smoother user experience.
Can I create my own voice commands for the Blazor Voice Assistant?
Yes, you can create your own voice commands! The Blazor Voice Assistant allows you to customize commands based on your needs. You can set up commands to control applications or retrieve information, making it tailored to your preferences.
What are the benefits of using a Blazor Voice Assistant in applications?
Using a Blazor Voice Assistant makes applications more interactive and user-friendly. It allows users to interact through voice, which is convenient, especially for hands-free tasks. This can enhance accessibility and improve overall engagement.
Are there any limitations to using Small Models and Voice Assistants?
Yes, there are some limitations. Small models may not handle complex tasks as well as larger models. Additionally, voice assistants depend on clear speech input. Background noise or unclear commands can cause misunderstandings or errors in responses.