Let’s GoLearn anything, in the openAboutDocsSign in
← Catalog

Docs / system notes

How this actually works

Read the material in the open. Enrol when you want the platform to remember the work you do.

01 / learner path

For learners

The short version

01ReadPublished lessons are open to browse.
02EnrolAttach progress to a published offering.
03SubmitSend work from an active enrolment.
04Keep scorePassed work is folded into lesson progress.

Reading and enrolment

Published course material can be opened without an account. Reading a lesson and seeing what an activity asks are separate from recording work. Enrolment is required before the service accepts a submission or writes lesson progress. Submitting work does require an active enrolment. Enrolment is only available for a published offering.

The four kinds of work

Supported activity types
TypeWhat the platform doesProgress effect
Multiple choice Checks the selected option ids against the author’s correct options in process. A fully correct answer can complete the activity and award its points once.
Written response Uses the author’s rubric when one exists. Marking checks that awarded evidence is quoted from the learner’s own answer; without a rubric, the attempt waits for a person. A fully marked answer can complete the activity and award its points once.
Project submission Records the submitted work against the artifact types requested by the author. It is left for a person to mark. Recording the attempt alone does not award automatic progress points.
Code run Supports starter code for runnable work. Language challenges run in the service’s sandbox; a challenge submission is checked by its verifier. A passing language exercise awards its points once; trying a run is not itself a progress award.

Progress and points

Progress is scoped to the learner’s active enrolment and lesson. A progress row keeps the completed activity ids and points for that lesson. Passing the same activity again does not add its points a second time. Submitting work or writing progress is refused when the enrolment is not active; re-enrolling reactivates the existing enrolment.

learner + enrolment + lesson
├─ completed activity ids
├─ lesson complete flag
└─ points earned so far

Practice challenges

The Challenges panel surfaces runnable language exercises from enrolled courses. Running code requires a signed-in session and checks the code without recording progress. Submitting an exercise requires an active enrolment; a passing verifier result records the exercise and awards its points once. Draft code is kept locally so an unfinished attempt can survive a refresh.

Offline

Installing Let’s Go puts it on your home screen and keeps lessons you have already opened readable without a connection. It does not fetch a whole course in advance, and it does not accept submitted work while you are offline: those still need a live connection to the service that marks or verifies them.

02 / author path

For authors

Course structure

Courseoutline + offerings
Modulesordered groups
Lessonsmarkdown + work

Writing a course

A course is an outline of modules and lessons, each lesson written in markdown, with any of the four supported activity types attached where work belongs: multiple choice, written response, project submission, or code run. Everything is edited in one workspace and saved as a whole, so a half-finished lesson does not go live by accident while you are still working on the next one.

Drafts and publishing

A course starts, and stays, a draft until somebody with the right to publish it says otherwise. A draft is visible only to its owner and whoever the owner has added as a co-editor; nobody else can see it, including in search or the catalog. Co-editors can write and revise everything in the course. Publishing and unpublishing are the owner’s call alone: a co-editor who tries either is refused by the server, not just hidden in the interface.

How marking works

Multiple choice is graded by a fixed rule: every correct option chosen, no incorrect ones, or it is marked wrong. There is no model involved and no judgement call.

A written answer with a rubric is graded asynchronously against the criteria the author wrote. A model proposes points per criterion, but the proposal is not trusted on its own: every awarded criterion must quote the exact words from the learner’s answer, and that quotation is checked in code. A submission with no rubric, and a project submission, is recorded for a person rather than silently marked.

Writing a rubric that marks well

Write each criterion as a specific, concrete claim a learner might actually put into words, not a quality you are hoping to sense in the writing. “Names the risk of storing the password in plain text” can be quoted against. “Shows a deep understanding of security” cannot: there is no particular sentence that satisfies it, so the evidence check has nothing to hold onto.

You may add a reference answer for the marker to read for context. It is never shown to the learner and the marker is told explicitly not to quote it: only the learner’s own words count as evidence for their own score.