My wife and I started the new year with a relaxing trip to the Central Coast of California—exploring the historic sites of San Luis Obispo and the stunning views at Pismo Beach. Before the trip, I decided to build a custom Gem to serve as our sidekick through it all. If you’re unfamiliar, Gems allow you to reuse prompts for your Google Gemini chats. You can also include files in a Gem’s knowledge base to better ground the chatbot. They’re similar to ChatGPT’s Custom GPTs.
The idea for the Gem was simple:
- Help plan a trip, either from scratch or by leveraging an existing guide.
- Serve as an on-demand resource throughout the trip, making recommendations and retrieving information with ease.
I turned to a good friend of mine who runs a travel blog. She’s a pro at crafting custom travel guides, and her sample guide for the Central Coast was the perfect foundation. She had previously created a tailored Miami guide for me in 2023 that I absolutely loved. With her Central Coast guide in hand, Gemini and I collaborated on an itinerary. It became my daily ritual to check in with Gemini for everything from activity reminders to menu suggestions. When I got to a restaurant on our list, I would ask it, “what’s the recommended dish to get here?” Each time, the Gem would pull directly from the guide, seamlessly blending my friend’s expertise with AI’s ability to synthesize and retrieve information instantly. It felt like the best of both worlds—human expertise paired with AI superpowers.
From this experience, I’ve developed a framework that will guide how I think about and approach consumer AI this year:
- Simplicity beats complexity. With AI advancing at breakneck speed, it’s tempting to get caught up in the race and feel left behind. I get it. People are already talking about agents. Sam Altman says OpenAI will achieve AGI this year. But maybe it’s time we take a step back. Let’s return to the basics and explore what we can still accomplish with existing, commercial-grade LLM chatbots. As Anthropic aptly states in their blog on building effective agents, “When building applications with LLMs, we recommend finding the simplest solution possible, and only increasing complexity when needed. This might mean not building agentic systems at all.”
- AI copilots can truly 10x your life. Travel With Me, as I named the Gem I built for the Central Coast trip, is just the beginning. I deliberately designed it to be open-ended, so I can reuse it for other trips this year. Beyond Travel With Me, I’ve already created FitCoach to support my half marathon training and Blog SideKick, my personal editor-in-chief for this blog. Building these tools isn’t just for tech enthusiasts or developers; anyone can do it. There’s room for all of us to create copilots that make life easier and more fulfilling.
- Human expertise remains essential. While there’s a rush in the AI industry to automate everything, I’m not here to optimize my life solely for the efficiency of it. Quality still matters to me. There’s little value in retrieving information quickly if it’s wrong or lacks depth. My trip was made extraordinary because of my friend’s guide. Likewise, my half marathon training is built on tailored advice from another friend who truly understands my goals and challenges. AI may be fast, but the depth and nuance of human input are still essential. This may change tomorrow but that is the case today.
I am a naturally curious person and there is so much more I want to explore with this technology. Gems and Custom GPTs are only one part of the equation. I hope sharing my experience inspires you to think about small pockets of your life experiences that could be made better with a touch of LLM magic—and that it encourages you to experiment boldly.
The potential is enormous, and as I refine these systems, I look forward to sharing my journey and maybe even releasing some of my Custom GPTs and Gemini Gems publicly.
Here’s to building in 2025!
Thanks to Chisom Okpala and my Custom GPT for reviewing drafts of this.