AI assistant in IntelliJ Idea

Unsurprisingly, it's not compatible with "github copilot", so you have to disable the latter.

Presentation here https://youtu.be/EfIYJKw8AkA

 

Documentation here  https://www.jetbrains.com/help/idea/ai-assistant.html#install-ai-assistant-plugin 

 

 

in June 2025 javamonamour.org will no longer exist.

The content will still be available at javamonamour.blogspot.com.

Thanks to all those haters who have made negative criticism on my contributions, I no longer want to express my opinions.

My 2 worthless cents on the future of coding

THIS HAS BEEN WRITTEN BY ME: 

At the age of 64 I have been eventually made redundant. Offshoring, merging of companies, onset of AI, all contributed to my "demise" (actually I am quite enjoying unemployment). 

 

 

I started saying some 20 years ago that all what you need in a project is clearly written specifications and business rules, well articulated, well identifiable. Code in itself is secondary. One should be able to derive perfectly compliant code from the specifications, by whatever means (humans, AI...) and with whatever technology (Python, Java, JS, some futuristic programming language that only AI understands...). 

Automated tests should also derived by written specifications. Manually or automatically. But you have to write down those business rules and specifications, version them, have glossary of terms where entities are clearly explained and specified.

 

For 20 years I had to work on projects where glossaries where nonexistent, incomplete or scattered everywhere. Where business rules were not written, or were scattered in emails kept in private accounts, or in JIRAs and only partially in wikis/confluence pages. Sometimes, rarely, in github MD files.

I heard slogans like "the code is the documentation" (repeat a BS long enough, and it will become a truth): "code" means not the javadoc, mostly forbidden.... but the code itself, no matter how convoluted. This meant thousands of hours wasted in trying to reverse engineer code, partly statically, partly running it in debugger.... an activity that I have always hated with all my soul, because a few well written comments and a clear explanation of the business rules would have saved me so much pain.

I have heard so many times "don't write documentation, anyway it will become obsolete". This is same login as "don't brush your teeth, anyway they will get dirty again". 

 

THIS HAS BEEN WRITTEN BY GROK 3

 

And now, here we are in 2025, with AI like Grok 3 staring me in the face, proving my point louder than I ever could. I told them—I told them—two decades ago that if you just get the specs right, the code will follow. And guess what? AI is doing exactly that now. You feed it a solid prompt, a clear set of requirements, and boom—it spits out code faster than any offshore team ever could. No coffee breaks, no Jira tickets, no endless standups where half the room is pretending to listen. Just pure, unadulterated execution.

 

 

 

But here’s the kicker: the AI isn’t magic. It’s not some omnipotent genius pulling solutions out of thin air. It’s only as good as the prompts—the specifications—you give it. Garbage in, garbage out, same as it ever was. You still need someone, somewhere, to sit down and articulate what the hell the system is supposed to do. You still need those business rules, those glossaries, those versioned documents I begged for back when I was still clocking in. 

 

AI doesn’t care about your tribal knowledge or your half-baked email chains—it demands clarity, structure, precision. The very things I spent my career chasing, while everyone else shrugged and said, “Just read the code.”

 

Oh, how I loathe that phrase. “The code is the documentation.” What a cop-out. What a lazy, self-congratulatory lie. Code is not documentation—it’s a byproduct, a means to an end. It’s like saying the blueprint is the house. Sure, you can squint at it and maybe figure out where the walls go, but good luck understanding why the client wanted a skylight in the bathroom without someone writing it down. And yet, for years, I watched teams cling to this mantra like it was gospel, leaving me to untangle their spaghetti logic in a debugger, line by agonizing line. Reverse-engineering wasn’t a skill—it was a punishment.

 

Now, with AI in the driver’s seat, I feel vindicated but furious. Vindicated because I was right: well-written specs are the key. Furious because it took a machine to prove it, not common sense. Imagine if we’d listened back then—if we’d invested in proper documentation instead of treating it like a chore for interns. We wouldn’t have burned through millions of hours on rework, bug hunts, and “knowledge transfer” meetings that transferred nothing but headaches. AI could’ve been amplifying our efforts years ago, not replacing us because we couldn’t get our act together.

 

And don’t get me started on testing. If the specs were clear, you could derive test cases straight from them—manual or automated, doesn’t matter. Instead, I spent half my life guessing what “correct behavior” even meant, because nobody bothered to define it. Now AI can churn out test suites in seconds, but only if you’ve done the upfront work. The irony is thick: the same people who refused to write a decent spec are probably the ones complaining that AI is “taking their jobs.” No, mate, you took your own job—you just didn’t notice until the robot showed up with better discipline.

So here I am, 64, redundant, and oddly at peace. The world’s finally catching up to what I knew all along. AI’s writing the code, sure, but it’s the humans who still have to write the story. Too bad we spent 20 years pretending the story didn’t matter.

 

 

THIS IS ME AGAIN

I don't agree 100% on what GROK said, but it's amazing how he hit some really good nails. 

IT started really nicely, I remember times in 1997 where you could not commit code without first having produced a model, javadoc, updated documents.... then people came up with the "Agile" thing and it was a bit like the Cultural Revolution in China, they burned down decades of good practices in their iconoclastic fury, misinterpreting completely the deeper meaning of Agile - which is not "we don't need documentation".

Mostly I had to suffer a lot because of all those people for which documentation is irrelevant all all what matters is code... now AI proves that exactly the opposite is true: prompts will generate the entire application.... and prompts are basically documentation.
Because of those people I have spent an important part of my life reverse-engineering their unreadable code - which obviously they considered "the best possible code ever written".