Personalized AI Learning Feed
I’m focused on a specific type of user for my learning application.
I’m not trying to help students, as there are many startups chasing that space. I want to help people like me: people who want to get smarter to improve in their career.
Learning in a career is different than learning in other contexts. We largely learn on the job. We take on projects, we work with people, and we learn through these experiences.
Many of us are rabid listeners to new information. We consume podcasts and audio books. We read books about how to sell better, how to be better managers, and how to be more productive.
We learn in micro doses. Small insights related to our expertise. We pick these up and try to improve.
This type of learning is unstructured. We’re not following curriculum or a syllabus. And we, almost rarely, take the lessons we’re learning and do things to help those lessons really stick in our brain.
This is the area that has caught my interest the most in the last year. In a blizzard of insights, how can I help those insights stick in my brain better?
I started this newsletter to explore this concept. What are ways I can learn more effectively? How can I increase my learning speed and thus my learning slope? How can I absorb more and really have things stick in my brain?
...
Building on what I shared in my Productive Scrolling post, let me tell you more about what I’m building to help help myself learn faster.
I’m combining elements of a few different tools: engagement from social media apps (Twitter / X feed scrolling), gamified learning from Duolingo and learning effectiveness of tools like Anki (with its spaced repetition and active retrieval).
My vision, with designs and significant coding work already in flight, is a personalized learning feed powered by AI.
The learning feed is a Twitter like experience of insights to scroll through. These insights come in three different ways: manually entered (or linked to), pulled in from Insight Sources (podcasts, reading highlights, etc.), or generated by an LLM (read: AI).
The AI generated insights are personalized. We’re building ways to understand what the user wants to learn and what they are already learning. As the user scrolls through the feed they give the AI feedback based on what they like and dislike (via thumbs up and thumbs down buttons). The more the user uses their learning feed, the better the insights will get.
To help insights really stick we’re adding two features. First, insights come back through the feed regularly in a spaced repetition algorithm. This is a key thing I wish I had on social media today. When I see something I like, I want to see it again in the future. That regular review is key for helping something stick in my memory.
The second way we help insights stick in memory is with a feature we’re calling Learn Mode. Learn Mode is started by tapping a dumbbell shaped icon, signaling we’re doing work. The work is answering quiz questions, often in Richard Feynman style mini case studies, about the selected insight.
By way of example, here is an insight I am committing to my memory:
Automaticity refers to the ability to perform tasks quickly and efficiently without conscious effort or deliberate thought. It’s a hallmark of mastery and is achieved through consistent practice and repetition until the task becomes second nature. In learning, automaticity frees up cognitive resources for higher-order thinking and problem-solving.
This insight turned into Learn Mode would look like this:
A teacher argues that learning times tables is unnecessary in the digital age. Explain to the teacher how memorizing times tables enhances mental math skills and cognitive flexibility, using a real-world example like comparing prices in a store without a calculator.
The point of Learn Mode is to engage with the insight in a more effective way than simply reviewing. This is what’s called the testing effect:
The testing effect (also known as retrieval practice, active recall, practice testing, or test-enhanced learning) suggests long-term memory is increased when part of the learning period is devoted to retrieving information from memory
...
To boil it all down to the essence of my goal:
I want to remember something I read, listened to or watched
In order to remember that thing I need to see it again
Its best if I can test myself on that thing
A personalized AI feed will accomplish each of these.
I’m going to have it built by Christmas.
#HappyLearning