Why we built Coursentials — and what we got wrong the first three times.
by Kingsley
When we started Coursentials, the brief was simple: take everything that made Storyline powerful, strip away the desktop install and the per-seat pricing, then layer modern AI on top so authors stop staring at empty slides.
We built v1 in a weekend. It was a beautiful disaster.
The boring parts are where the value lives
We optimised for the demo — generate a course from a prompt in 90 seconds. That worked. What didn't work was everything around it: the SCORM packager rejected by Moodle, the LTI launch that 404'd in Canvas, the narration cues that fell out of sync after a re-render.
The first three rewrites were about the unglamorous middle: a content- addressed publish pipeline, an LTI 1.3 helper that handled deep linking + AGS + dynamic registration as first-class citizens, a schema validator that catches malformed scenes before publish.
AI is a multiplier, not a substitute
The wizard generates a starting point. The author edits. We learned to make every AI step skippable, reversible, and clearly marked. Authors trust AI when it stays in its lane.
That's it for now. We have a lot to ship through the rest of the year — multi-language generation, cmi5, XLIFF — and we'll write about each one as it lands.