RIP RubberDuck

A year ago, I set out to find my ikigai. I found it, but not in the way I expected.

tldr:

• I'm shutting down RubberDuck
• I'm really proud of what I built
• I learned a lot, but it's time to move on
• You can still find my educational content on YouTube

Here's how the journey went.

First, Thanks

I can't start this without thanking everyone who supported me along the way.

Specifically, I want to thank my wife for supporting me in my quixoticism, and always encouraging me to follow my dreams.

Countless people helped me over the last year, and I'm grateful for all of you.

I'm also grateful for the users who tried out RubberDuck, and have continued to use it. I'm sorry I couldn't keep the lights on for you.

Vision

The most exhilarating part of the last year was being able to chase my own vision. I set out to build a platform to help engineers learn faster with the Duolingo model. You can see the original pitch and how I was thinking at the time.

I was solving my own frustrations with learning new software engineering concepts (Rust at the time), and combined it with my love and efficacy of the Duolingo model. I remain convinced that this is the future of learning, and that the market was validated by the success of Duolingo and other learning platforms.

Learning to pivot

Comitting to a vision is important, but so is knowing when to pivot. The hardest part for me was letting my ego go and accepting what data was telling me. This has been the biggest lesson I've learned in the past year.

Lesson 1: Let go of my ego

A few months in, I had shipped the software-engineering-specific learning platform, but didn't have enough data to tell if I had good product-market fit. I moved into marketing mode, and created some surprisingly viral videos on Nix which brought in a lot of users. But the retention was low, and I was burning through my runway.

I made 2 pivots I regret, because they were confusing for users:

1. The platform wasn't getting enough traffic. I chased SEO and traffic by creating a newsletter. This was a mistake, as it was a completely different product. I should have focused on retention instead of acquisition. It turns out nobody wants another newsletter in their inbox 😅
2. The newsletter wasn't working. Based on the success of the YouTube videos I created, I decided to pivot the platform to just be video-based courses. This model could work if I had a larger audience, but I found it hard to convert visitors to purchase the course. There are also a lot of really good competitors in this area

Measure twice, cut once

The common thread in both of these pivots was that I didn't have enough data to make informed decisions. I should have spent more time talking to users, and understanding what they needed.

Lesson 2: Talk to users earlier

This led to my final pivot, which was the most successful. I interviewed as many users as would talk to me, and the original interface was what they were all the most excited about, and got the most value from. I returned the product to the Duolingo-style interface.

But I still had a retention problem to solve - users would churn once they learned what they came for. Could I increase retention with different content? Maybe if there were enough different types of content, users would stick around longer.

Around this time, I played around with generating a math course - and it worked! With the help of GPT4, RubberDuck expanded to "Duolingo for everything". This led me to an interesting thesis I plan to write about someday: LLMs are lossy compressed knowledge, and can be decompressed into whatever interface you want.

Building a faster horse

It turned out that the expanded content actually helped acquisition more than retention. But the users I interviewed would still take the course that they wanted and bounce. I wasn't solving a struggle for them, otherwise they would have stuck around.

This was the next lesson I learned: I talked to a lot of users, but I didn't ask the right questions. Instead of asking what they wanted, I should have asked what they were struggling with. This would have given me a better idea of what to build.

Lesson 3: Solve a struggle

I had one more theory to increase retention. My intuition was that the mobile web experience was easy to forget, felt clunky, and that a mobile app would increase retention in several ways:

1. sending push notifications
2. a more native experience
3. a wider audience in the App Store
4. be top of mind for users (home screen)

I was able to ship the app in record time, and it was a great experience.

Unfortunately the app didn't move the needle on retention either. At this point, the only remaining conclusion was that the product wasn't solving a struggle for users, and I wasn't going to find product-market fit with this approach.

I had two more ideas:

1. People had fun learning with RubberDuck. I could double down on the entertainment aspect of the app, introducing a "trivia mode" with the now-generated content. This would unfortunately put me in competition with some well-funded competitors.
2. I'd been avoiding a sales cycle by targeting a consumer app. Instead, I could sell the interface as a Learning Management System to educational institutions. This would be a completely different business, and I didn't really have the energy to start over. After some research into this path, I found that most users of LMSs are locked into long-term contracts, and that the market is very competitive.

I decided to move on.

Life gets in the way

Starting a 0 to 1 business is incredibly difficult, and I didn't appreciate the effort it would take along the way. I ended up spending a lot of time on things that weren't building revenue or helping users learn faster, like:

• interviewing users
• filing taxes
• filing a trademark
• dealing with App Store requirements
• complying with GDPR requests

The beauracracy burned me out, and I lost the passion I had when I started.

Lesson 4: If you want to go fast, go alone. If you want to go far, go together.

Education is still broken

I applied for several startup incubators and seed funds, but nothing came of it. In the process of building a deck though, I found some concerning statistics about the state of education:

• Every year, over 1.2 million students drop out of high school
• Around 40% of undergraduate students leave universities and colleges every year
• In 2021, the poverty rate for high school dropouts was 27%, while the poverty rate for high school graduates was just 16%

This is in the United States alone, despite spending over $1 TRILLION on education every year!

I still believe that the Duolingo model is the future of learning, and that there is a huge opportunity to improve education for so many people.

What's next?

I'm proud of what I built, but I'm even more proud that I tried at all.

I took a few weeks off to reflect on what I learned, and what I want to do next.

I baked a lot of bread!

During this time, I realized that I missed building things with other people (see Lesson 4). Working by yourself gets old... I interviewed with a few companies, and found a great fit at DitchCarbon.

DitchCarbon uses machine learning to extract emissions data from sustainability reports. Companies can use this data to reduce their scope 3 emissions by better understanding their supply chains, and choosing more sustainable suppliers.

Ikigai

When I started RubberDuck, I wanted to build a business making the world a better place by helping people learn faster.

DitchCarbon provides a way for me to try making the world a better place in a different way - and potentially at a much larger scale. By the end of my tenure, I hope to be able to calculate the CO₂e I've saved by helping companies reduce their emissions.

On top of that, I get to work with a team of incredibly talented people, and learn from them every day. The problem space is fascinating, and I'm excited to concentrate my technical focus on machine learning!

Happy Earth Day 🌍

Cam