June 6, 2026, 11:19 p.m.

When I began the 'MS Oracle' section in my website, I added a simple 'Chatterbot' bot to it, it was extremly basic with limited data; it was actually trained via YAML files with data I add manually, so you can imagine the 'tiny scope' of the bot -_-.

Every new conversation felt like a test I had already written the answers for. If a visitor asked something slightly outside my handcrafted YAML patterns—say, a niche historical date or a philosophical take on a current event—the bot would either freeze, spit out a default “I don’t understand,” or loop back to a generic greeting.

It was, to put it kindly, a digital parrot with performance anxiety.

I spent weeks fine-tuning those YAML intents. I’d add ten new question-answer pairs, feel proud for an hour, then watch someone ask a question that should have been covered but wasn’t—because they phrased it differently. The maintenance alone became a second job.

Then came the shift.

After months of studying late into the night—reading papers, breaking demo projects, and rebuilding broken agent architectures—I finally moved from a static chatterbot to Agentic AI.

Now, instead of one fragile bot, I have a small team of agents connected to LLMs. Each agent has a role: one Agent is called "Dr. Mary" for English speakers and the other is "Dr. Heya" for Arabic speakers; you can talk to them from here:

The scope of the produced bot(s) exploded, of course, to cover almost any topic I want to focus on.

Suddenly, my “MS Oracle” could talk about everything from ancient prophecy methods to modern machine learning bias—without me writing a single new YAML file. It can summarize recent articles I’ve linked, compare ideas across different sections of my site, and even admit when it doesn’t know something (rather than guessing poorly).

The difference isn’t just technical; it’s philosophical. A chatterbot stores answers. An agentic system constructs them.

And the best part? I still control the guardrails. The agents run on my chosen LLMs, follow my custom instructions, and stay within the topics I want to focus on. But now, “focus on” means a universe of possibilities rather than a prison of pre-written lines.

If you’re still debugging YAML files or watching your simple bot embarrass itself in front of users, take the leap. Study agentic design. Let your bot fail spectacularly with an LLM—then watch it learn to fly.

My YAML files are now archived. The chatterbot sleeps. And the Oracle? It’s finally worth asking a real question.