Three Things That Changed in NStrim This Year: AI Module Authoring, Multilingual Sessions and AI Marking

Teaching time is the scarcest resource in any classroom, and for most of the last decade a depressing share of it went to installing tools rather than writing code. NStrim has always handled that part — a personal subdomain, Git repository and database for every student in under sixty seconds. This year we went after the hours on either side of the lesson: drafting the module, delivering it in every student's language, and marking what comes back.

NStrim AI: describe a topic, get a full module

Writing a good practical task is slow work. You need a brief that is unambiguous, a framework choice that suits the group, a starter repository that compiles, and assessment criteria you can defend to a moderator. NStrim AI now drafts all four from a description of the topic.

It produces the full task description, recommends a stack — Laravel and PostgreSQL, say — scaffolds the starter repository, and writes pass/fail criteria and a rubric. What it does not do is publish anything on your behalf. You review the draft, change what you disagree with, and publish in one click. The judgement about what your students should be learning stays where it belongs.

Multilingual sessions: author once, read in 15+ languages

Vocational classrooms are rarely monolingual, and a student who misreads the brief fails a task they could have completed. Modules authored in NStrim are now delivered to each student in their own language, across more than fifteen of them, from a single authored source.

The teacher writes once. Every student opens the task in the language they think in. For international competitions this removes an entire category of unfairness that used to be handled with printed translations and crossed fingers.

AI marking, with the teacher in the loop

Marking thirty submissions consistently is harder than marking one carefully. AI marking runs the rubric you already defined when you authored the module, flags the edge cases that need a human eye, and gives you a starting score for each submission.

Three things make it defensible. It marks against the criteria you wrote, not criteria it invented. Snapshots let you replay any moment of any student's work, so a contested grade can be examined rather than argued about. And every score is a suggestion until you confirm it — the platform never finalises a grade on its own.

The rest of the stack is where you left it

Everything underneath is unchanged and still included: Docker containers isolated per student with reusable image templates, Gitea repositories and a browser-based editor, PostgreSQL and MySQL with pgAdmin and phpMyAdmin, and more than thirteen frameworks ready to go. Mission Control still lets you pause the whole room, add ten minutes mid-exam, push a corrected file to everyone, and take snapshots for rollback and grading checkpoints. And it still runs EU-hosted, on-premise, or delivered on-site.

If you would like to see any of this against your own curriculum, book a demo — or try the live demo without signing up for anything.