HEX Seedbed
A development and design training program that took over a hundred people from zero to their first real project. The experiment was treating teaching as a product problem.
- Role
- Program design & mentoring
- Year
- 2024
- Status
- Archived
- Stack
- React · Flutter · Figma · Discord · Notion
The Seedbed started as an excuse: we wanted to hire good juniors and couldn't find them. Instead of complaining about the market, we built the program we wished we'd had.
It ended up being the most instructive thing I've done on how people actually learn, and why most free technical content fails.
The hypothesis
Online courses have terrible completion rates — somewhere between 3% and 15% depending on who's counting — and the usual explanation is "people lack discipline". That explanation is comfortable and it's wrong.
Our hypothesis was different: courses fail because they optimize for syllabus coverage, not for the moment something works. A student who writes forty isolated exercises and never sees their code running in front of another person has no emotional reason to come back on Tuesday.
How it was structured
Six parallel tracks — programming fundamentals, UI/UX levels 1 and 2, Android, Flutter and frontend — sharing one calendar and one rule: they all ended on the same Friday, with everyone presenting in the same room.
That produced something we hadn't planned for. The design people started understanding the developers' constraints, and vice versa. By the third cycle, final projects were cross-disciplinary without us asking.
The numbers
A 68% completion rate against the usual 3–15% isn't explained by better content. Our content was, frankly, worse than many paid courses. It's explained by Friday's demo.
What didn't work
- The premium tier. Charging for "extra mentoring" created two classes of student and poisoned the group dynamic. We dropped it in cycle three.
- Discord as the main platform. Excellent for live conversation, terrible as memory. Anything resolved in a thread was lost by the following week and someone would ask it again.
- Unstructured one-on-one mentoring. With no agenda, sessions drifted into therapy. Occasionally useful, but not what we'd promised.
Why I archived it
Because it worked at human scale and I couldn't find a way to grow it without breaking the one thing that made it work: that I knew every participant's name and project.
That's probably the most expensive lesson in the whole thing. Some good things don't scale, and that's fine. What isn't fine is pretending they do and degrading them in the attempt.