.NET Rocks!

Is it time to build your own agent harness? Carl and Richard talk to Emmz Rendle about her work on Daemonic AI, which gives you more control over which models and tooling you use to build software with agents. Emmz talks about the upcoming rug pull in AI software development tools, where prices are rising, and services are being restricted. Having enough control to choose when to run locally becomes key to being productive at a reasonable price. Being able to pick-and-choose what agents and configurations to use for each of the agent roles you want to implement is super powerful - check out the GitHub project and take it for a spin!

What settings, configurations, and workflows do you use for every .NET app? Carl and Richard talk to Scott Sauber about his list - from organizing folders by feature, to logging, security, and testing. Scott talks about enforcing rules like treating warnings as errors so you won't ignore important warnings, and validation in the build, to make applications more reliable. Each of these items represents some work, but in the end, your application will be higher quality and more reliable. Which ones are you already doing?

Large Language Models can generate a lot of text - but is it any good? Carl and Richard talk to Vishwas Lele about his ongoing efforts at pWin.ai to build tools for responding to government RFPs. Vishwas focuses on the quality problem - both the quality of the incoming RFP and the quality of the responding proposal. How do you determine the key requirements of an RFP reliably? And when it comes to the response, how do you provide measurable results for a response? The conversation digs into a change in workflow that benefits the RFP process regardless of tooling - and gives hints to the patterns of success with LLMs!

Use What Works! Carl and Richard talk to Dylan Beattie about the Use What Works movement, encouraging developers to use well-maintained open-source projects available today rather than rolling their own. Dylan explains how folks go down a path of believing a library is simple until they learn enough to realize that every bit of software is more complicated than they realize. And the less code you own, the happier and more productive you are. Adding AI to the mix only makes it clearer: you need some stability in development. If you're changing every layer of code, you'll spend even more time and frustration chasing problems. Make getting results easier - use what works!

Ready to go nano? Carl and Richard talk to José Simões about the open source .NET nanoFramework - a community-driven project to provide .NET for embedded systems. José talks about the evolution from the .NET microFramework, to something even smaller, while at the same time, microcontrollers have gotten much more powerful. The conversation looks beyond the hobbyist and educational uses of these systems into commercial IoT applications. The development cycle is one you'll recognize, working in Visual Studio (or Visual Studio Code) and executing against an emulator, or to the actual controller via USB. And yes, you can set breakpoint in the controller!