Live Demo https://obnexus.vercel.app/
GitHub https://github.com/Isaac-aiai/obnexus-project
Not Ready Yet? Good.
I got handed an internship project right out of school: build an AI scheduling assistant for an OB/GYN ward. My immediate reaction was something like panic. I could barely write SQL without looking things up. How was I supposed to build this?
But here's the thing — the scaffolding was already there. Beautiful landing page, complete chat interface, even those little suggested question buttons. It all looked real. Except the AI was fake. Every response was hardcoded.
It was like a team with jerseys printed, stadium rented, fans in the seats — but no players on the field yet.
My job was to make this team actually play.
Why Agent and Not Just a Chatbot
Before writing any code, I had to figure out why we needed an Agent at all. Why not just wire up a chatbot and call it done?
This is actually a critical question for anyone building AI products, and I've been thinking about it a lot.
A chatbot, the way I see it, is like a midfielder who can only pass — never shoot. You tell it to press the opponent, it says "sorry, my configuration is possession-based." You tell it to take a shot, it says "I'm just a passing specialist."
An Agent is different. Agent = brain + legs + the autonomy to decide when to pass and when to shoot.
You say "find me empty beds," and it figures out which table to query, what conditions to use, how to structure the results. It doesn't just understand your intent — it acts on it.
That's the kind of player who can actually score.
For someone like me who wanted to quickly validate whether AI could do something useful, choosing Agent over Chatbot felt like choosing a versatile striker over someone who can only juggle — I needed someone who could put the ball in the net.
The Learning Strategy I'm Betting On
Here's a counterintuitive take: don't wait until you're ready.
The default learning approach is sequential — watch all the tutorials, solve all the exercises, feel "prepared," then finally start building.
It's like an athlete saying "I'll play my first competitive game after my free throw percentage hits 95%."
The problem is — you'll never be ready. The pressure of real competition, the rhythm of actual opponents, the coordination with real teammates — training can't simulate any of that. You only learn where you're weak by actually playing.
I decided on a different strategy:
Find something the 12-months-from-now version of yourself might be able to complete. Start now.
obnexus became my real game:
- Complex enough to be worth the effort
- Clear milestones where I could see progress
- Tech stack I actually wanted to learn (Agents, databases, cloud deployment)
- Most importantly — after this game, I'd have game tape to show interviewers
What's Next
I've got the team structure. Now I need to make them actually playable:
- Wake up the AI (right now it's just a mannequin)
- Connect the database (so it can see real game data)
- Implement tool calling (so it can take actions, not just watch)
- Deploy to production (so the audience can see the game)
Each step is a small goal. Add them up, you've got a complete match.
A Thought
Here's what I keep coming back to:
You don't become skilled by being fully prepared. You become skilled by repeatedly stepping onto the field.
When I graduated, I thought I needed a few more years of study before I could tackle challenging projects. Now I realize the fastest growth comes from taking on something beyond your current ability — and learning as you go.
The version of me three months from now will definitely be stronger than the current one.
That's why I'm willing to step onto the field now. I don't know exactly how this will go, but I figure that's kind of the point.
Top comments (0)