Navigating the Idea Maze of Better Ways to Learn
The idea maze is a concept introduce to me by a 2013 Chris Dixon essay. The concept is that there are many possible paths an entrepreneur travels to refine an initial idea. There is rarely a single eureka moment where a brilliant startup idea is discovered. A startup founder travels a market space, exploring the trade offs and the multiple paths that could be trodden.
I’m at the entrance to the idea maze today with my vision for an AI powered learning system. Notice I’m dropping the description of “journal” as I’m finding it misleading in early user interviews.
My starting insight is summarized in my observations as a busy knowledge worker:
Learning quickly is a huge part of the job
Most of my learning activities are highly inefficient (listening to podcasts, reading books, watching lectures)
Creating rigor / effectiveness around my learning process is a highly manual effort
The core problem I want to solve for the world is to automate learning; create a system where learning and memory retention happen automatically with AI.
There are so many tools out there to manually create a learning system.
Anki - open source, flashcard based spaced repetition system that is incredibly popular. You can download sets of flashcards to teach yourself most anything.
Obsidian - open source, knowledge based graph with a plethora of plugins to piece together a learning workflow.
Roam Research - knowledge based graph database popular among academics to organize what you are learning, search through it easier, etc.
Mochi.cards - an improved UI take on Anki that looks really interesting.
What has caught my attention the most the last few months is that LLM technology is capable of automating most of what is occurring in manual learning systems today. An LLM can take information that the learner wants to learn and create a set of quizzes to help the learner master the material.
I did this myself for my learning goals. I wanted to learn more about AI, so I read a few articles then asked the LLM to quiz me on those articles (quizzing is a reliable way to learn more effectively). I created my own spaced repetition system where daily I would re-quiz myself on things I was learning.
The result, with this simple prototype, was outstanding for me. I learned more about AI (and then expanded into other topics). I could expound on embeddings, RAG, supervised vs. unsupervised learning, etc. I mastered concepts I was roughly familiar with such as SaaS business metrics.
I believe a quiz based learning system is a superior learning system to most of the ways my peers are learning today. Quizzes, especially when they are spaced out over time, are the key activity LLMs can bring to a knowledge workers learning system to make it much more effective.
My product idea is simple:
Given a set of learning information (blogs, podcasts, etc.)
Create a set of quizzes to help the learner remember the information
Spread out those quizzes over time
Adapt the quizzes dynamically to the progress the learner is making (if they are knocking the quizzes out of the park, then make them harder and vice versa)
...
Back to the idea maze concept. Where do I take this set of insights and a rough product idea?
A knowledgeable investor told met here are three ways to make money in EdTech. You can sell direct to consumer, to education institutions or to corporations’ HR department. I’ll use these three as the initial possibility routes:
Maze Pathway: D2C
Direct to consumer has some significant pros and a plethora of cons. The market is potentially massive if (and this is a big if) I can figure out a way to reach consumers with a willingness to pay. The market has some interesting dynamics though. For example, how many people love learning for learning’s sake? This group of autodidacts may be much smaller than a broader market, which may be OK depending on the size of company I’m trying to build.
The con is this is the most price sensitive group. How much would you someone pay for a learning tool to help you remember things better? I’m still figuring this out but I can’t imagine it’s more than a few dollars per month (at least in my current version of the idea). I’m not convinced the idea I have is not going to be best as a free, open source tool like many of the other tools in this space. If open source is the right route for this tool, then there may be a different business model to explore.
I’ll of course continue to explore this pathway as it fits closest with my initial vision of how I could use the tool myself. My exploration tactic is to spend significant time on social media trying to see if my ideas can catch those that would be interested.
Maze Pathway: Education Institutions
This route has some interesting possibilities to it and was from at least one investor conversation a non starter.
On the one hand I can see why investors wouldn’t love this. The sales cycle has got to be ginormous. I haven’t sold here, but I can’t imagine these are quick sales.
On the other hand, I can see the flip side of a long sales cycle: sticky product. I imagine once you’re in, you’re in for a while. I’ll need to get much smarter here though before I can explore far. This is an area of EdTech I know the least because I am not an educator.
I’m not building a company for investors (at least yet as I’m self funding) so I’m not entirely scared to approach this. However, I do want to learn from the wisdom and pattern recognition of others so I’ll tread carefully.
My exploration tactic is to network and meet people in the space. Today for lunch I met a friend who taught me about courseware subscription businesses. One of the takeaways is there is demand for an exam prep tool with high adaptability. There may be a route in a tool for companies already selling successfully to education institutions.
Maze Pathway: B2B
The investor I talked to liked this pathway through the maze the best and it was the route her company ended up taking.
B2B education purchases often are done through HR but they can show up in other departments. Sales leaders have a need to educate their teams. Engineering leaders want their engineers to be up to speed on the latest tech.
There are many examples of large B2B edtech companies in my own backyard. Pluralsight is a large Utah company that’s been able to pull off this playbook well (with, from what I can tell, also a playbook to sell to consumers directly).
I like this route and will continue to explore it. Similar to the Education Institution play there may be a route to sell a white labeled tool directly to the service providers already in the marketplace. For example, I could take the course ware adaptable quiz idea and create an LLM specific to creating quizzes for course content creators. This could then be sold in a B2B context by other service providers where we sit as a white labeled tool in their platform.
....
“Oh, the Place You’ll Go”
The fundamental purpose of a startup (and why I define my company as a startup even absent venture funding or, currently, venture ambition) is the search for a reason to exist. Is there a business to be built around a spaced repetition quizzing system powered by AI?
I’m not sure but I’m committed to find out.
Some random thoughts as I read:
-I’m an obsidian user, and while part of the whole point is to let your brain think in a divergent way and make connections in a free-flowing way, I’ve always wished there was a way to integrate the whole body of knowledge better.
-It strikes me that a big sticking point may be how to get the information that you want to quiz into the model in the first place. Distilling a large body of knowledge into what is the most important aspect to quiz on is probably what takes up most of my time. Providing some sort of lubricant at this step would be most advantageous in my eyes.
-Investors would prefer the B2B model because that’s seemingly the quickest way to scaled profitability. From an education institution perspective, Qualtrics obviously comes to mind as a company that was wildly successful that started in this space. I think the big advantage their product had was that it appealed to researchers (the core of the education space where the $$$ is) but could also be used for other groups. I’m wondering how much utility researchers would see in a product like the one you envision - where qualtrics helped them do the thing they are incentivized to do, the learning is (sadly) only a secondary goal for most academics.