AI and the Developer

People often ask me for killer ideas on what "AI" solutions they can pitch to their clients. Honestly? That's not as simple as it sounds. Why? Because the real question isn't about AI itself—it's about the value it can deliver. And since I'm neither a "client" nor a subject matter expert in every conceivable business domain, chiming into this discussion can feel like giving cooking advice without ever tasting the dish.

However, there's one glaringly obvious gap that's surprisingly overlooked: how we, as software engineers and architects, can harness AI to dramatically improve our own workflows. Yeah, I know—we've all laughed at the memes where someone asks ChatGPT to write their code, only to spend hours debugging gibberish. That's not a problem with AI—that's just reality reinforcing the age-old adage: garbage in, garbage out.

But what if someone competent—someone who genuinely understands software architecture—leverages the rapidly advancing capabilities of AI tools? Then, dear friends and based on my own experience, we're suddenly capable of breaking the sound barrier and delivering breathtaking results at unprecedented speeds. And no, AI won't replace the architect or engineer. At least not the good ones. Not the ones who know exactly what they want to build, how to use the tools properly, and when to doubt the AI's outputs.

Because, trust me, when everything aligns: boy, oh boy, have I been amazed.

And this is exactly where we should focus: not just on "how to bring AI (projects) to clients" but "how to use AI FOR the client." For example, to save costs or deliver higher quality. It doesn't always have to be the next Uber.

What are my daily learnings as a software engineer? I can't claim these are complete—just a snapshot.

First: keep an overview of all the new tools and frameworks coming out almost daily. Balance openness to new things with maintaining consistency. I, for example, currently work more and more with Cursor in the Microsoft ecosystem, but I'm not ditching VS Code. Both have their justification (especially when it comes to extensions and DevContainers, there are still good reasons for VS Code).

Second: some subscriptions are truly worth it, and you should invest in them. I started with a paid version of OpenAI's ChatGPT (which I still have), tried GitHub Co-Pilot (which I don't currently use), and moved just recently to a Cursor subscription—which is so far worth every penny, in my opinion.

Next point: who thinks AI really "understands" what we want, or can independently write entire applications—well, not quite. Or maybe on a "Hello World" demo level. But that's not what we want.

During a PoC for a potential major re-platforming project, I had my first experiences. At first, it's overwhelming what Cursor spits out: whole projects, dozens if not hundreds of files. But under the hood? A Frankenstein collection from heterogeneous sources—impressive to look at, but hardly useful.

Break down tasks cleanly and logically, set precise contexts, proceed step-by-step—and suddenly magic happens. Entire (sub-)modules are proposed cleanly and require only minor adjustments (either via prompt or manually).

And especially for somewhat annoying, time-consuming tasks, AI is a genius helper: test preparation (no, I still wouldn't trust AI to fully create proper tests), checking in code including perfect commit descriptions and comments, creating or adapting CI pipelines, and so on.

None of it is rocket science—but that's exactly the point: as of today, we can use AI to significantly accelerate our development process—and, if applied correctly, improve quality.

For example, when checking in, I sometimes (😉) didn't put much effort into writing detailed descriptions. Or when creating a pull request, I forgot to mention which tests were run. But—well—if the AI knows exactly what was done and generates a structured, clean summary for me, free of charge—thank you very much!

---

Bonus Tool Tip:

Take a look at WARP Terminal. I was never a big fan of the command line. But that has changed. This tool is smart, intuitive, and makes me want to control everything via the CLI. Give it a try.

---

In short: AI won't replace us anytime soon—but used intelligently, it allows us to build faster, smarter, and—frankly—just a lot more enjoyably. Who knew coding could be fun again?