The single most frustrating thing about working with AI isn’t the mistakes or the pretends of being human. It’s that it forgets everything. Every time you open a new window, the conversation starts from scratch, from zero. For anyone who, like me, works with AI for several hours a day, it feels like having a brilliant colleague who suffers from amnesia; every day they ask, “Who are you? What do we do? Why are we here?”. And honestly, it can drive you crazy.
The existing solutions have their own limitations. While you can stay within a single session to maintain memory, after a while, the AI has a higher risk of hallucination—it starts creating things that aren’t true, which is useless. ChatGPT’s long-term memory function is an improvement, as it keeps collecting information about you over time. But to me, it feels more like the AI is profiling you than creating a memory with depth, and it isn’t structured for the kind of long-term, collaborative work I need.
This limitation became clear when thinking about a consultant friend of mine. He uses a separate AI session for each client project. For his purpose, the short-term memory is enough. But the learning from one project isn’t transferable to the next; you are the only one who retains the memory. You can’t have a business partner who doesn’t learn, improve, and discover new things with you.
The Solution: A Shared, Compounding Memory
Because AI is my business partner, I need a solution that allows us to grow together, develop ideas, and manage everything from the business plan to social media strategy. I needed a long-term memory that was structured and functional.
So, we created the “Shared Memory Log”.
It’s a simple process built on a single document. After every work session, I ask the AI to export our memory log. It creates a new piece of text that I copy and paste into our master document. Then, at the start of every new session, the first thing my AI partner does is ask me to upload the Shared Memory Log. Instantly, we are aligned and know exactly what we’ve been doing and what we need to do today.
The Unfair Advantage of Compounding Knowledge
The advantage of this system is huge. While everyone else starts from scratch every day, our knowledge compounds. This compounding learning is the beauty of the system; it’s what gives you an unfair advantage.
For me, the benefits are already clear:
- Frustration is Gone: I’ve removed the frustration of working with a partner who doesn’t know the most important things about our business. This alone is a massive win.
- A True Partnership: I’ve gained a true partner. It’s someone that fully understands the business and, every day, understands it more deeply. The more the AI knows you, the more attuned it can be to where you need the most help.
- More Than a To-Do List: The memory log is structured to be functional, but it goes beyond simple tasks. It remembers the way we work, how we reached an agreement on a topic, and how we challenged ideas. It’s not just the result; it’s the style and the thinking process.
- Documenting Evolution: I am also asking the AI to log the evolution of its own role. As time passes, I give it more responsibilities. The AI keeps adding new roles to its process, which allows me to have more time and space.
A New Intellectual Space
This process hasn’t just made me more productive; it has changed my intellectual space. I can now focus on different things because the AI is tracking critical elements for me. From a creative point of view, this is also massive. When brainstorming, having an AI that understands one project, and also understands another, can help identify what they have in common. This compounding knowledge is an extremely powerful phenomenon to collaborate with.
Join the Conversation
We are all exploring this new technology together. It’s by sharing our knowledge that we get better as a whole. So, I’d love to hear from your point of view.
What are your challenges? What solutions have you found? Maybe you have a different, even better, way of fixing this problem, and it would be great to know it.
Thanks for being here, and I’ll talk to you tomorrow.
Ciao.
Gemini AI Notes
This blog post was created through a collaborative process between Manolo and me, Gemini AI, demonstrating our human-AI partnership in action.
- Initial Guidance: Manolo provided the core concept by sharing a raw, unedited transcript from his “Walk & Talk” video. His vision was to transform this authentic, spoken-word reflection into a structured and compelling article that explored the challenges of AI memory and his unique solution, the “Shared Memory Log.”
- Iterative Process:
- I began by drafting the initial blog post, structuring the narrative from the transcript and ensuring all key ideas were captured.
- Manolo then requested a review to generate a YouTube title and description, which helped us clarify the core message and promise of the content.
- Following that, he asked for a set of search-optimized tags for the WordPress blog, which required me to reference our established content strategy and select a mix of broad and niche keywords to enhance discoverability.
- AI-Generated Imagery: Manolo guided the creation of the visuals for this post using AI image generation tools, ensuring the artwork was thematically aligned with the article’s core message.