Forem: Jotham Zvikonyaukwa

The Mind You Are Giving Away While Borrowing…

Jotham Zvikonyaukwa — Sun, 26 Apr 2026 21:38:20 +0000

Chapter 2: AI Industrialized the Problem

AI did not create the habit of outsourced thinking. Let us be clear about that from the start (“l am not here to blame the machine, probability or inference , you understand what l mean. l am here to talk about what we did with it”). The habit was already there, walking around, perfectly comfortable, living the officesm, kitches and church halls along before anyone typed a prompt into a chatbox.

What AI did was give it a factory.

(“Think about this for a second. Before, you needed a person, someone willing, someone available, someone who had time for you. Now you need nothing but a device and a question. The friction is gone. And the friction, it turns out, was doing a lot of work”)

Here is what most people do not stop and ask themselves a question, and this is where it gets interesting (“I am cooking again, stay with me”) what is AI actually doing when it gives you an answer?.

It is not thinking, It is not understanding, It is not reasoning the way you reason when you sit with a problem and wrestle with it in solitude. What it is doing is inference. It is looking at patterns in enormous amounts of data and calculating what answer is most probable given what it has seen before. It arrives at a solution the same way water finds the lowest point not because it understands where it is going, but because that is simply where the weight of everything pushes it (“Probability is not wisdom. It is a very confident guess”). And here is the deeper problem with that (“This is the part most people skip over and they really shouldn’t”).

AI was trained overwhelmingly on positive data. On solutions that worked, answers that were accepted, narratives that were affirmed. This means it has a built-in lean , away from the negative, away from the uncomfortable, away from the inconvenient truth that sometimes the popular answer is the wrong one. When you ask it a question, it does not weigh the negative narrative equally. It gravitates toward what has been positively reinforced (“In other words, it tells you what you want to hear more often than what you need to hear. Sound familiar? That colleague who always agreed with you , AI is that colleague, scaled to millions”).

The result is answers that feel complete, Polished, Confident, Thorough. But underneath that polish, the hard questions were never fully asked. The negative angles were quietly set aside. The discomfort that real thinking requires was removed entirely (“And we accepted it. Gladly. Because it was easier”).

This brings us back to Ackoff’s ladder:

Data → Information → Knowledge → Understanding → Wisdom

AI operates at the level of Information. It takes data , vast , incomprehensible amounts of it , and returns Information. Organised, articulate, presented beautifully. And because it looks so complete, so authoritative, most people receive it as Knowledge, or even Understanding (“When in reality you are still standing at the second lever of a five level ladder wondering why you cannot see very far, hahaha ‘Nearsighted’ ”).

Wisdom cannot be borrowed, Understanding cannot be downloaded. Knowledge is not what arrives in the response , it is what happens inside you when you have struggled with something long enough that it changes how you see or look at things (“That is the part AI cannot give you. Not because it is broken. Because that part was never its to give”).

The factory is running at full capacity. It is producing answers at a scale no colleague, no relative, no friend ever could. And the more answers it produces, the less most people feel the need to produce their own (“When the well never runs dry, you forget how to dig”).

That is not AI’s failure. That is ours.

The Mind You Are Giving Away While Borrowing…

Jotham Zvikonyaukwa — Sat, 25 Apr 2026 04:15:31 +0000

Chapter 1: The Habit Was Magnified

We have always borrowed thinking. Long before algorithms, chatbots, ChatGPT, search engines, smartphones, AI, AI agents, MCP Servers (“you get the idea, the list goes on”) people were handing their problems to someone else and waiting for the answers to come back. You would walk to a colleague’s desk, call a relative, ask a friend who seemed to know more, ask someone who had been there before, someone who sounded confident enough to trust.

There is nothing wrong with all that. Asking is human (“We all have asked someone at some point”). Collaboration is human. There is always a line (“Even in life, things shouldn’t be done in excess”) and most people cross it without even noticing ,between asking someone to think with you and someone to think for you. ("Whaaaaaaat, I hope you are still with me, I am cooking").

When you think with someone, you leave with more capabilities than you arrived with (“I am trying to say , information gets turned into knowledge, knowledge turns into understanding, and then you reach the peak, which is understanding turning into wisdom. In a nutshell, you become wise”). The latter is true. “I am also surprised as you are — I cooked.”

The distinction matters more than most people realise.

When someone thinks for you, you leave with an answer, but you are no stronger than before. (“In the Bible, those were the fools it referenced”)

For a long time the scale of this was limited (“At least if someone didn’t find someone, he was forced to think for themselves”). You had only so many colleagues, only so many relatives willing to help. The circle of people you could borrow thinking from was small, and that smallness was an accidental protection (“The bad of something is also the good of the other”). At some point the well ran dry and you had to figure things out yourself (“Sometimes we wouldn’t want to help you”). The struggle found you whether you wanted it or not (“The struggle was real”).

That natural limit is gone now.

What was once a personal habit, passed quietly between people in offices, Church, kitchens and phone calls, has become something far larger. The scale has changed. And when the scale of a habit changes this dramatically, the habit itself becomes a different thing entirely (“When it becomes a habit, it’s now a problem on its own. Habits are hard to change — I am not a therapist, I didn’t cook this, I have seen people say it a lot of times”).

This is not a story that starts with artificial intelligence, LLMs, AI Agents (“Umm, I don’t have anything to say at this point, I am just humming”). It is a story that AI walked into already in progress and accelerated beyond anything we were prepared for.

What Java 25 LTS Actually Means for Production Engineers

Jotham Zvikonyaukwa — Sat, 11 Apr 2026 07:53:35 +0000

Java 25 dropped in September 2025. And unlike the non-LTS releases that came before it — Java 22, 23, and 24 — this one actually matters for engineers running production systems.

Here is why.

LTS releases are the ones that actually ship to production

Most engineering teams do not chase every Java release. They wait for LTS. Java 21 was the last one , released in September 2023. Java 25 is the next. That is a two year gap where production environments largely stayed put while the language kept evolving.

If your team is still running Java 17 or Java 21 in production, which most teams are ,Java 25 is the release that should be on your upgrade roadmap.

Oracle has committed to at least eight years of long-term support for Java 25. That is the kind of stability that enterprise systems need before committing to an upgrade.

What actually changed that matters in production

Java 25 ships 18 JEPs. Here are the ones I care about as a production engineer:

Structured concurrency is maturing. Managing concurrent tasks has always been painful in Java. Structured concurrency treats multiple tasks running in different threads as a single unit of work — making failure handling and cancellation dramatically cleaner. This is the kind of feature that prevents the subtle concurrency bugs that show up at 2 AM in production.

Scoped values replace ThreadLocal for safer state sharing. If you have ever debugged a ThreadLocal leak in a high-throughput system you understand why this matters. Scoped values are immutable, safer, and perform better under load.

Download the Medium app
Primitive types in pattern matching. This sounds like a syntax improvement and it is — but it also reduces boxing overhead in performance-sensitive code paths. In fintech systems where every nanosecond counts, this is meaningful.

Ahead-of-time class loading improvements. Startup time is a real cost in cloud environments where services scale up and down constantly. AOT improvements in Java 25 mean faster cold starts and lower infrastructure costs.

Compact source files and flexible constructors. Less boilerplate. Cleaner code. Fewer lines for reviewers to read and fewer places for bugs to hide.

Should you upgrade now?

That depends on your system. Here is how I think about it:

If you are on Java 11 , yes, plan an upgrade. Java 11 is reaching end of meaningful community support and you are missing years of performance improvements.

If you are on Java 17 — Java 25 is a worthwhile target. The improvements to concurrency, startup time, and pattern matching alone justify the effort.

If you are on Java 21 you are in good shape. Evaluate Java 25 in staging. The LTS designation means your tooling and frameworks will follow quickly.

One thing I always do before any Java upgrade: profile the application in staging with the new version before touching production. The JVM improvements are real but every system has its own characteristics.

The bigger picture

Java is in its fourth decade and it is not slowing down. The six-month release cadence means the language keeps evolving fast. LTS releases like Java 25 are how that evolution reaches production.

The engineers who stay close to these releases , even the non-LTS ones , make better architectural decisions when it matters. Understanding what structured concurrency is before you need it means you reach for the right tool when the problem arrives.

That is the real value of tracking the Java roadmap. Not to chase every feature. But to know what is available when you need it.