Skip to main content
Tessera keeps a decision log. Accepted decisions are constraints: changing one takes a new decision that supersedes it, not an edit to the old one.

​
The ones that shape every screen

One course → module → lesson → block hierarchy everywhere. Grading shows the submission and rubric side by side, and a unified Today list is the entry point. AI features edit this structure and never create a parallel one.
Every AI output is a labeled draft with its source, a diff, and Accept, Revert, and Regenerate. A person must mark it kept before learners see it. Agentic bulk actions appear as previewable change sets. The tutor gives hints and worked parallel examples, not answers, unless policy allows.
Personalize on goals, role, time, prior knowledge, language and reading level, accessibility, device, and bandwidth. Offer every format to everyone, and show every adaptation with a reason and an Undo.
Administrators set limits per program; instructors choose a mode per activity within them, and locked options say why. The answer key is never a tutor source. Instructors see per-learner summaries; full transcripts open only from a submission.
Phase 2 builds a React component app with shared design tokens and Storybook. It’s designed to grow into a full-stack app on the same Cloudflare deployment. See architecture.