࿊ On Agentic Problem Solving

The framework for breaking down and solving problems in the age of AI is learned through repetition, but here are my top tips for succeeding in the space today.

1. Come at problems with an open mind and a willingness to experiment. The tools allow you to break down a problem into various subproblems that you can reason about until you understand them from first principles. An example of first-principles thinking with an AI agent is this exchange:
   • “How do I go about doing X (in this case, we can use the example of connecting from my iPhone to my iMac computer)? Tell me the different pieces that are needed. Respond in one paragraph and one sentence.”
   • {AI explains}
   • You ask follow-up questions about the AI’s explanation and parts of it that are new to you.
   • Once you have a goal in place, you move on to step 2.
2. Coding / software development from first principles.

Arguably, you do both of these at the same time. With enough repetition and projects, you will have an understanding of how this technology interacts with your mind. Examples of recent projects I have embarked on:

1. A generator nameplate extraction tool
2. A battery revenue estimator
3. A list building, enrichment, and outreach system
4. A personal finance alert system
5. A NAICS code classification engine
6. A demand response & virtual power plant enrollment figure engine
7. A long-running persistent skeptic agent for code review
8. A deep research prediction market system
9. Load shedding visualization
10. Stealth project

I hope to continue learning more as I interact with more stakeholders who look for solution engineers and AI-native creatives.

All the best,
Leonardo A. Cruz