Per-learner journey narrative

A running per-learner narrative the platform keeps up to date as the user works. Every touchpoint completion (chat3 turn complete, stage submission pass, note-checker attempt complete, kickoff understanding complete, learn-mode session end, cowork session end, presentation end) writes one entry into the journey. The entries compound into one moving picture for each focus area: what the user understood, what they had to be walked through, what they re-applied successfully later, where the same gap keeps showing up.

Most platforms read a learner one event at a time. To coach someone well, the platform has to know the running story across all their work: not just whether they passed this stage, but whether they passed it the same way they keep passing or the same way they keep stalling. The journey map is that story. The next manager turn reads it, the next heavy recalibration reads it, the end-of-scenario review reads it. Without it, every event re-reads the user from scratch and the manager forgets what it just learned about them.

One centralized handoff table per learner, append-only across touchpoints. Every touchpoint completion writes a handoff entry: what just happened, what about the user this signals, what should carry forward to the next event. The write happens after the touchpoint's own evaluation lands, so the entry is grounded in a real verdict rather than a guess. Downstream readers (the heavy-completion calibration, the end-of-scenario review, the manager NPC team) pull the last N entries for the focus to ground their next move. The compounding shape is what makes it work: each entry references the prior one, so the story stays coherent across sessions instead of resetting every time the user comes back.

Tech lead. Owned the compounding-handoff design, the per-touchpoint write schema, and the read paths the rest of the platform uses to ground its next decision.