-
Notifications
You must be signed in to change notification settings - Fork 3
Devlog
Translated from Spanish by 🌐💬 Aphra.
Published May 25, 2025
New project! This time about learning and teaching using Artificial Intelligence.
Imagine you rely on an AI model to study for months, but each conversation starts from scratch. The model helps you... but doesn't remember that you already master algebra but get stuck with trigonometry.
When you search about Artificial Intelligence applied to education, you find the same thing everywhere: prompt templates, automation of activity creation (and answers to those activities), automatic corrections (yes, but not), the search for the perfect GPT... Old formulas for a completely new paradigm. Students looking for how Artificial Intelligence can do all their tasks and teachers looking for exactly the same thing, but at the same time wanting students not to be able to do it. I already wrote a blog entry a couple of years ago hinting at the dangers of this approach (https://davidlms.com/en/docentes-enfundad-vuestras-armas-contra-la-ia/), so I'm not going to repeat myself.
And two years later, we're still lacking sufficient debate on the topic. How we use these new tools to improve learning and how we integrate them so that students are as well prepared as possible to succeed in this new paradigm. The answers aren't easy at all. We'll have to try, make mistakes, iterate through solutions, and keep searching for perfection in that which cannot be perfect.
In the midst of all this confusion, today I'm releasing the LearnMCP-xAPI. What does it consist of? It allows an Artificial Intelligence agent to add and query learning records. Think of it as the Memory concept of a language model, but instead of storing user preferences and facts to take into account in future interactions, it would store what the user knows and doesn't know about the elements they're learning. A person could use it to set learning objectives and have the language model adapt to their progress. It could also be used by teachers to obtain assessment evidence tied to specific evaluation standards from student interactions with AI tutors or similar.
I know it will be difficult for many people to get a clear idea of the utility of this project. That's why in the coming days I'll share some use cases with visual examples.
LearnMCP-xAPI doesn't claim to be the solution to anything, just a component, a piece of the puzzle that we can design to adapt education to this new reality, keeping it free and universally accessible.
What if AI could really be your long-term learning companion, not just your tutor of the moment?