Builders
#253 - May.2025
The future belongs to the builders!
Anyone with curiosity and creative judgment can now build solutions with a few clicks.
This weekend I experimented with creating AI agents. The experience left me both excited and unsettled.
Excited because the possibilities feel endless.
Unsettled because I realized that this is a big paradigm shift.
But I guess many people have the same feelings. The good news is this is just getting started!
Here are some quick/initial learnings from my experiment creating agents:
- Abstraction. Today's technology is abstracting complex code as never before. Sure, you can still dive deep (if you want to), but is not needed. This democratizes building but creates a dangerous knowledge gap.
- Creator block. The new paradigm of agents is transformational, in such a way, that I felt creative block: "how can I used this technology?". Is intimidating to think about all the possibilities. This is what AI is enabling today.
- Momentum + Community. Everyone is experimenting right now. There's a big momentum on exploring how AI can help us build new things. You can feel the excitement when looking into the different tech communities. People are very supporting providing guidance on how to fix technical problems.
Start small. Experiment. Pick one specific problem you face daily. Build an agent to solve it. Learn from that experience.
What are you waiting? Go and build something.
¡Saludos!
👨🏻💻
For people interested in tech datapoints from this experiment:
→ I'm using a local version of n8n (is free and you can run it through Docker)
→ For quick prototyping I installed Ollama model locally. There are tradeoffs (there are better models tailored for type of tasks) but I optimized for speed.
→ Serpi (Google Search API) gives you a credit of 100 searches per month. Enough for testing.
→ In short, for prototyping and experimenting, I spent 0$.
→ I'm using a local version of n8n (is free and you can run it through Docker)
→ For quick prototyping I installed Ollama model locally. There are tradeoffs (there are better models tailored for type of tasks) but I optimized for speed.
→ Serpi (Google Search API) gives you a credit of 100 searches per month. Enough for testing.
→ In short, for prototyping and experimenting, I spent 0$.
