AI doesn't make typos, I do (and I'm learning): My raw notes from the Google Agents Challenge

Back in November,I dove into the "5-Day Gen AI Intensive" with Google and Kaggle. I entered as a Visual Artist and Consultant, looking for tools. I left with something different: a new mindset about our human role in the AI era.

They say that to truly learn something, you have to write it down by hand. So, I unplugged the keyboard and took out my notebook.

The evidence of the crime (and the learning)

(My real notes during the course. No filters. Yes, they are in Spanish).

If you zoom in, you might spot something curious (besides the Spanish): spelling mistakes.

I could have asked an AI to transcribe this and polish it before publishing. But I decided to leave it "raw" for an important reason:

I am aware of my mistakes, and I work on them every day. At my age, I keep learning, improving my writing, and challenging my own neuroplasticity. But this image demonstrates the perfect division of labor:

My human brain was busy connecting abstract concepts, designing strategies, and visualizing complex architectures.

The AI is there to assist me with syntax and final execution.

This is the essence of Augmented Creativity: it's not about being perfect, it's about orchestrating machine intelligence with human intention.

3 Insights from my notes (translating my own handwriting)
Looking at my scribbles, I rescue three technical concepts that change how we use AI at PixelIA LAB:

1. "They are machines for predicting the next word, not databases"
   It's easy to forget this. Interacting with models like Gemini on Vertex AI reminded me that if we don't give them "grounding," they hallucinate. They don't "know" things; they predict things. Understanding this technical limitation is the first step to stop using AI as a magic oracle and start using it as a logical inference engine.

2. The Era of "Context Engineering"
   In my notes, I wrote in big letters: "(Ingeniería de contexto)". As a consultant, I see that the future isn't just prompting (asking for things). It's designing the environment where the agent lives. What documents can it read? What tools can it touch? We are the architects of its memory and its boundaries. Without good human context, the most powerful model is useless.

1. From Chatbot to "Flow Orchestrator" What blew my mind the most (visible in my notes on "complex flows & multi-agents") is that we are no longer talking about a Q&A chat. We are talking about systems where Agent A investigates, passes the info to Agent B who drafts, and Agent C critiques.

Conclusion: The Human Orchestrator
This Google and Kaggle course confirmed to me that technology moves fast, but human skills (critical thinking, curiosity, and the ability to learn regardless of age) are steering the ship.

AI can correct my spelling in milliseconds, but the strategic vision and empathy to apply these solutions to real problems... that's still on me (and you).

What about you? Do you still see AI as a chat, or have you started orchestrating your agents?

Written by Mongui Rogers - Visual Artist & AI Consultant at PixelIA LAB.