Field note · December 16, 2025
Year three
The course is finished. What that took, and what we would do differently.
Thirty-six lessons across ten modules. Seven tools. Fourteen datasets. Twelve articles. The course is done.
The original estimate was six weeks. It took three years, which is a ratio we intend to bring up for a long time.
What took the longest: not writing. Rewriting. First drafts take about two hours a lesson. Making one good takes six more, and the work is almost entirely subtraction — deciding what to leave out and saying so explicitly instead of covering everything badly.
What we would do differently: build the tools first. Three visualization lessons were written, then the chart builder was built, then all three were rewritten — because building the thing taught us what the lesson should say. The same happened in miniature with the SQL module and the playground.
What surprised us: the appetite is for the unglamorous parts. Pipelines, quality, "how do I know it actually ran". The statistics modules get read; the quality module gets used, and the broken dataset outdraws everything.
What we got right: free, and the licence. People have translated lessons, put modules in company onboarding, and forked the whole site to make one about a different subject. None of that needed our permission, which was the point.
What is next: the course being finished does not mean it is correct. Corrections go in the page with a note about what changed, rather than being silently edited. We will keep writing field notes, and there are a few articles queued.
Some of those are scheduled well into next year, which the build handles by simply refusing to publish a page whose date has not arrived. If you are reading this later, there will be posts below this one that did not exist when it was written.