Citations by default, code that actually runs, current events as raw material. A workshop you can hand to a class without lowering the ceiling for advanced students.
asks a question, drops a reading, or opens a code assignment.
routes through Learn, Research, Search, News and Code as needed.
every claim links back to its source — students can follow the trail, teachers can grade it.
Today · what you can do
Not a sequence. Take the one the lesson needs and leave the rest.
Bring lectures, slides and readings in from Canvas, Brightspace, Moodle or Blackboard, and Learn builds summaries, quizzes, flashcards and narrated recaps from the real material — not from what a model half-remembers about the topic.
Hold a conversation that cites every claim it makes. Good practice for the student, useful for the teacher reviewing the trail.
Branch and version a Research notebook for next week's lesson, with sources attached and merges when two threads converge.
Hand out a Code project where students pair with the coding agent — it works alongside, not instead of, the learner.
Build the week's reading from deduped, multi-source news clusters. The world becomes the textbook, with citations.
Reading lists, drafts and rubrics live in versioned folders the rest of the workshop can search semantically.
Today · in private access
Each carries its live status, so nothing on the roadmap is presented as available.
What the reading list is built from
Alongside whatever you import yourself. Each of these is a reference page you can read before you set it as reading.
What a programme actually teaches, read from the institution rather than a directory's summary of it.
Product manuals and book-length texts read page by page — including the ones published only as documents.
Open-access research read in full — with the datasets a finding rests on kept alongside the write-up.
The settled account of a subject, read from the maintained entry rather than a copy of it.
Students working on their own projects have their own page: AI for student research.
Shapes of work
Sketch a term plan from a learning objective. The Writer drafts the unit outline, the Researcher attaches readings to each lesson, the Summariser builds the one-page recap a substitute can teach from. Editable, versioned, ready for the department head review.
Sit beside a student working on their independent project. The workshop questions before it answers — pushing for the source, the assumption, the alternative reading — and never claims a fact it can't cite. The trail of work is the assessment artefact.
A class of thirty, each on their own thread, each at their own pace. The Tutor agent holds the line on rigour — Socratic where it should be, direct where the student is stuck — while progress analytics show who has stalled and where.
Drop a submission against a rubric. The workshop suggests a band, marks the rationale next to each criterion and flags passages it can't fairly judge. A first pass that respects the teacher's final call — never a hidden one.
One workflow, end-to-end
Four steps across the workshop, each shipping today.
Skim the week's deduped multi-source news clusters. The same story read four ways collapses into one — exactly the teaching moment you want when the syllabus says 'evaluate sources'.
Pull the cluster into a Research notebook. Branch a thread for each angle — economic, ethical, technical. Citations are first-class, so the case study models the citation discipline you want from students.
Open a Code project. Pair with the coding agent to turn the underlying numbers into a chart students can read in five seconds. The agent runs the formatting passes alongside; you keep the editorial call.
Export the case study and the chart into the term's versioned Storage folder. Next year's class starts from the artefact, not from a fresh search — and the version history shows what you changed.
Not available yet
Drop a file, ask a question, get a chart. No formulas required.
Questions
The workshop is Socratic by design — it asks for the source, pushes for the alternative reading, and works alongside the student rather than instead of them. The trail of the conversation is visible and makes a useful assessment artefact in itself.
Yes, for course material. Learn imports courses from Canvas, Brightspace, Moodle and Blackboard — lectures, slides and readings are read page by page and become the source the study tools are built from. Artefacts made elsewhere in the workshop still export from Storage and are uploaded back manually; there is no grade write-back.
Learn writes a timed exam from your own course material, grades it on the server, applies browser-level proctoring and gives an AI debrief on every attempt. Progress analytics build from the real attempts, so weak topics and exam readiness come from what a student actually did rather than from self-report.
Both. Teachers use Research and Storage to build lesson artefacts, and Learn to turn a course into study and assessment tools. Students work on their own threads at their own pace, each with their own progress record.
For education teams
We'll walk a department through the rooms that are open today and what a class actually does in them.
