AI drafts. Professors decide.
Tallam is built on a small number of commitments. They shape every screen, and they are the reason some things ship later than a demo would suggest. Open any of them to see where it shows up.
- 01
Every AI change is a proposal with a before and after. Approval applies it and records a revision. There is no path where the assistant edits a course by itself.
In the product: the course editor shows the current text beside the proposed text, with Approve and Reject as the only ways forward. Revision history lists what changed and when.
- 02
Assistance draws on the materials in the course. The tutor being built will answer only from professor-approved sources, with citations, so students learn the course and not the internet.
- 03
Reading annotations belong to the reader. Chats are personal. Roles and course access are enforced on the server, not in the browser.
- 04
Available, in progress and planned mean exactly that. Demonstrations are labelled as illustrations. Nothing counts as done until it is verified on the public site.
- 05
The goal is students who build judgment, ask better questions and can tell a good answer from a fluent one. Fewer clicks and clearer pages exist to make space for that.
How a course comes together.
Upload the syllabus, readings and notes you already have.
Originals are kept and downloadable. Text is extracted for review but the file is never altered.