I’ve often talked about how much fun it is for me to work on small software projects in my “spare” time, and over time some of these projects have become small pieces of software deployed for customers or made freely available to the communities (mainly Citrix and Nutanix).
A few months ago I discovered a way of developing that was new to me and that has turned the way I approach development upside down; this new approach involves using Visual Studio Code, integrated with Copilot Pro+ and the various AI agents (in particular Claude Sonnet 4 and 4.5).
AI is certainly not my discovery…these days the web talks about nothing else; so a couple of months ago I decided that the time had come for me too to try developing with the help of AI.
At first my approach was cautious and skeptical…my previous experiences with AI had been interesting, yes, but limited to ChatGPT-style chatbots (mostly the free tiers).
Once I made up my mind to try Copilot Pro+, my experience changed COMPLETELY.
I discovered how convenient it is to develop with Visual Studio Code using the agent mode of the various AI models, which let me truly appreciate the benefits of having AI at hand, from handing it the job of analyzing code to hunt for bugs, all the way to asking it to write entire functions and important pieces of code completely (or largely) on its own.
Unlike my first experiences with AI, where I just had it analyze pieces of code I had written so it could debug them, agent mode now completely changes the approach, letting the human focus on thinking about what they want to build and leaving much of the code writing to the AI.
I realize this approach can be “scary”, since the inherent risk is that developers no longer know the code they have developed (or maybe DIDN’T develop??), but the result is a truly mind-blowing paradigm shift, especially for people like me who don’t do development as their main job.
Over these months I’ve been able to take on projects in programming languages I didn’t know (and who would have found the time to study them??), and I’ve built small pieces of software (web ones too) that I would never have been able to build on my own; all with mind-blowing simplicity, with nothing more than the will to do it and clear ideas in my head.
In the past I often found myself held back by my limited development skills (again, I do it in my “spare time”) and I was left with ideas “in my head” that I couldn’t “turn into reality”; today there is a way to get past these difficulties! I’m not saying it’s a cure-all, or that serious developers won’t be needed anymore, but people like me who have ideas and want to try to make them happen can now give it a go! I’ll leave it to those who are more of a developer than I am to judge the quality of the code I’ve written this way; but code that, on my own, I (in several cases) wouldn’t have been able to write at all (neither well written nor badly written).
Oh, one more thing: have you ever tried hooking AI up to a server to have it figure out why the server is misbehaving? It sounds surreal, but AI can understand logs and error events at a speed that we humans will NEVER be able to match…and the result? You can now solve problems much faster!! Here too, I’m not saying sysadmins (the role that best describes me) won’t be needed anymore, but a sysadmin who has access to AI (and knows what to ask it and how to interpret its answers) can do their job at unmatched speed.
Relax! Don’t think I’m crazy…I’m not handing my customers’ information over to AI…I would NEVER allow myself to connect my customers’ servers to public AIs (like ChatGPT, Claude, etc…) because I know very well what that would mean; but when I was able to connect it to my own lab servers, it helped me solve problems that would otherwise have cost me a lot more time.
As I said above, I don’t see it as a cure-all, but as an IMPORTANT opportunity to raise the level of us humans, by handing the less “brainy” tasks over to AI…and what’s wrong with that? Nothing, I think.