Three conversations on AI and coding: what they mean for Kentico developers

I listen to a lot of podcasts. My preferred app, Pocketcasts, says since 2017 I've listened to them for 107 days and 21 hours. I subscribe to many, but most focus on software development and digital marketing.

Three that I've subscribed to for years have recently featured discussions about AI and the impact it's having on software development. These episodes broadened my perspective and echoed my mix of excitement and concern about the uncertain future of the field.

Let's explore some of the ideas they cover and what they could mean for developers building Xperience by Kentico solutions.

(Links to all the podcasts can be found below. I recommend listening to all of them 🙂.)

Hanselminutes: Is Vibe Coding Real? with James Montemagno

If you've never heard of Scott Hanselman, the host of Hanselminutes, you're either very young and new to this industry or so old you're already forgetting things. His guest on this episode, James Montemagno, works at Microsoft supporting developer communities and has been writing C# since the mid-2000s. Both know their tech and software development but they sit on different sides of the AI fence.

Scott is cautious about vibe-coding, especially taking it all the way to production. James, however, sees it differently, based on his experience shipping a sentiment analysis app built largely with AI assistance:

• Total Code: ~17,000 lines

• Code He Wrote: Only 20 lines (complex JSON parsing that AI couldn't handle)

• Timeline: Zero to deployed in 3 days

• Tech Stack: Blazor frontend, Azure Functions backend, Azure AI services

Here's how James explained vibe-coding:

If you apply the same concept of how we've been developing software to working closely with our AI counterparts to help us get that code completed, then to me that is the vibe. Now the traditional vibe coding definition is forget the code, the code doesn't exist. Be one with the ai. That's not my vibe, Scott. And I think everybody has their own interpretation of vibe coding.

[...]

So as I've been evolving, I've been vibing as the AI does something that I like, I update my instructions to tell it to go do that. [..] If it does something I don't like doing, I tell it don't to do that [..] So as I'm working with it, I am, for all intents and purposes, crafting and making it better over time.

[...]

if I already have a project or I've been working in a project and vibing in a project and I've been working on these best practices, it's gonna follow my best practices.

James emphasizes building a personal AI toolkit:

• MCP (Model Context Protocol) servers for extended capabilities

• Multiple AI models for different tasks (Claude for exploration, GPT-4 for focused work)

• Integration tools like Perplexity for research, GitHub for issue creation

• Custom workflows tailored to individual development preferences

And, there are specific areas where he is much more trusting with "vibing" out his code, like CSS and design and initial code generation.

This honestly feels like the approach I've been taking recently - craft a prompt, supply context, see what the AI generated. If it doesn't match what I'm looking for, I adjust the context (ex: with rules files). I also rely on the existing code that's been generated by AI and approved by me or I've crafted myself. This context also helps keep things on track.

Yes, there are a variety of challenges that come with allowing the agent to do more of the work, like generating code you don't truly understand or code that will become a maintenance challenge, but these existed before AI.

• If understanding the generated code is important for security and predictable behavior, developers can use AI to learn how the code works and become more knowledgeable in the process!

• If maintainable software is a project priority, developers have tools (instructions and context) to focus the agent and produce maintainable code.

James shifts perspective by highlighting other ways agentic software development benefits developers:

• Removing toil from development work (like updating Dancing Goat to the new content retriever API)

• Enabling projects that wouldn't otherwise get completed

• Learning opportunities in unfamiliar technologies

• Focus on interesting problems rather than boilerplate

Whether or not you think vibe-coding is real, or the future, or whatever... explore, try stuff out, and find your vibe with AI tools.

The Changelog: Agent, take the wheel

In this episode, the Changelog crew talks with Thorsten Ball, creator of Amp - an agentic coding tool - and an enthusiastic advocate for the power it brings to developers.

Thorsten wrote a blog post titled How to Build an Agent or: The Emperor Has No Clothes which walks developers step by step through the non-magic process of creating an AI agent. He feels that developers should look what's going on behind the scenes with the AI tools being marketed to them. If they see there's no magic, they will gain an appreciation and a deeper understanding that will help them adopt and be more productive with agents.

a lot of the copy from other AI software is this “AI, magic… It knows everything about you… It’s going to replace you, and it’s going to replace your job, and whatever you’re doing.” And for me, the fascinating bit is that these are incredibly powerful tools. Let’s figure out how to yield them, let’s figure out how to make real use of them, and just… Come along, let’s use this. Everything is changing. These are incredible tools, that will change software in the next years tremendously.

When a programming language feature doesn't behave how you expect, or a component library doesn't produce the right user experience, developers typically investigate a layer deeper, read the source code, and explore how the technology works. This helps them use their tools well and come up with creative solutions.

AI technology has been moving at such a rapid pace that most developers haven't had time to explore how things work. This leaves us in a position where:

• AI is everywhere.

• Adoption is expected.

• Developers see it fail when pushed beyond demo scenarios.

In other software development technology spaces we would pause, read the source, and understand why.

Thorsten suggests that we can and should do this with AI as well. He mentions the example of agent's tool calling - it's actually not that complex!

[...] with tool calling, you basically start a conversation with the LLM and you say “In the following conversation, when you feel the need to, say, read a file, or list files, or…” What else? “Run a terminal command, respond in this specific way. Respond with a message that starts with “tool call, name, read file”, like in a specific syntax. And they’re trained on this. So when the model “thinks”, in quotes, air quotes, for everybody listening, if it thinks it needs to call a tool, it will respond in a specific way. And that’s it. That’s the whole magic trick.

[...]

And this has collapsed into the model, because now you can give them tools, and they do this on their own. And it’s truly like just a for loop.

And the funny thing is if, I don’t know, if you ask a hundred engineers, half of them would say “It’s just a for loop.” And the others would say, with a smile on their face, “It’s just a for loop. Like, this is crazy.”

So, if developers did take the time to explore the software behind AI agents, they would realize they're surprisingly simple at their core:

• Tool Calling: Give the AI access to simple tools (read file, list directory, run terminal commands)

• Loop Until Success: The agent tries approaches, gets feedback, and iterates until the problem is solved

• Minimal Code: A functional agent can be built in ~300 lines of code

Additionally, something you tried a year ago might have failed or required too much effort, and that experience sticks with you and leads to valid skepticism. But, because AI technology is moving so quickly that experience and skepticism might already be outdated!

Here are some reasons developers remain skeptical about AI, as mentioned in the podcast:

• Outdated Experience: People tried early AI tools and dismissed them

• Unrealistic Expectations: Expecting AGI-level performance immediately

• Learning Curve Denial: Treating AI as magic rather than a tool requiring skill

For me, this episode was a great reminder that two things can both be true.

• AI agents are powerful

• AI agents are simple

CoRecursive: When AI Codes, What’s Left for me?

CoRecursive is one of my favorite software development podcasts because Adam, the host, does a great job featuring guests that have their own unique stories. In this episode, he shares his own story about becoming interesting in building software, eventually entering the industry, and how he feels AI is changing things.

Although he enjoyed typical childhood activities, he was most excited about building things (emphasis mine):

[...] I was hunched over my computer, building a mastermind board game in Turbo Pascal. And you know, I would go outside and I would go play. But the most exciting things for me were those building moments. The building stuff was when I had the time of my life.

I've been focused on web development exclusively since 2012, learning all the technology and coding skills that have been key to working in this industry. But, if I think back to when I was much younger, my first interest related to building software was when I discovered I could build custom maps in the original Starcraft.

I used triggers, timers, counters, and branching logic to create stories, characters, and experiences others could play. I was a game developer first and what I really loved was building. No programming languages, no code, no package management, just tools that helped me create something that others could enjoy.

Adam lived through the past 10-15 years of exciting and growth in web and software development in general, but now he's seeing things really shift in a whole new direction with AI:

I feel like I’m pretty pragmatic and somewhere in the middle, but I’m definitely leaning towards excitement after, I don’t know, 20 years of coding for a living. I’ve seen enough tool changes to recognize when something’s worth exploring. I feel like if you love building things, if you like to create, nobody can take that away from you. Not a boss, not a new tool, not even yourself. The urge to build these side projects and experiments that I love to create, it’s just part of who I am, right? And AI can’t take that away from me, but maybe it can help me.

This quote is relevant to my story at least, because I learned to enjoy building with technology before I had all the tools I've become productive with, so I believe I can continue to enjoy building with a new set of tools - including AI and agents.

But that doesn't mean I'm comfortable with this change, probably because there's some fear involved.

I think that a lot of resistance to AI coding tools comes from the same place: fear of losing something that has defined you for so long.

So, two forces are at play here - fear of losing control or being replaced, and inflated expectations fueled by AI hype. Billions are being invested in technology and products products... and marketing. You can't escape it!

Thankfully, Adam also gives some advice on how to overcome this fear.

Treat it as a chance to learn something new. Start small, right?

[...]

But these tools aren’t magic. Like they’re not gonna nail every problem on the first try. I think of it more of like working with a teammate. You need to go back and forth. You need to refine what you’re working on. You need to collaborate to get good results. [...] That’s not a flaw, that’s just how building things actually works. Right? The same would happen if I were designing something with a colleague.

Here's the summary of the approach I've been taking:

• No biases - ignore the hype and leave your skepticism behind.

• Adjust your expectations - start with something small and give this new technology an honest try... it's not magic!

• Remember you enjoy building - code is just a tool we use to build; there are other tools and we have many other skills we use daily.

• Don't forget you're still allowed to code - you can balance which skills you keep sharp and how much time you spend on them.

• You're going to grow and adapt and be ok 😉

So, where do we go from here?

If you're a software developer, building Xperience by Kentico projects with .NET, C#, React, and the classic web platform technologies, the AI technology wave is impacting you. Agentic AI tools are here and helping some people experience real capability and efficiency gains. You could be one of those people! I could be one of those people!

We're all in the same boat together. The boat isn’t leaking or sinking - it’s ready to ride the AI wave to new and exciting places.

At Kentico we're working to help you have a better experience adopting AI into your solution building workflows.

• We're releasing an official MCP server, to give AI agents direct access to Xperience by Kentico documentation and guides, reducing confusion with past Kentico products.

• We plan to release more tools for agentic development workflows, helping you build and update Xperience projects more efficiently and adopt new features faster.

In the meantime, check out articles on the Kentico Community Portal about AI technology and agentic software development. And be sure to share your experiences with us in the discussions for this post.