Introducing Kernex: Rebuilding Coding Infrastructure in Education

One of the First AI Code Detection Systems for Coursework. Period.

Most coding platforms in education still treat “academic integrity” as a checkbox: run a plagiarism tool, maybe flag a few copy-pastes, and call it a day. AI-generated code changed the game, and almost nobody in the classroom space had an answer. We did. Kernex ships KernexCheck 1—the model behind our AI code detection—inside KGrader, one of the first production systems built specifically for student submissions. We don’t just guess; we use research-backed methods and combine them with line-level feedback so instructors see where in a file the model thinks AI might have been involved, not just a single number.

How We Do It (Without Getting Too in the Weeds)

We care that our AI signal is explainable and defensible. KernexCheck 1 is trained and evaluated on datasets like AIGCodeSet and others; we're moving to our own classroom code sets so we can keep improving on real assignments. Under the hood we combine two signals: a trained classifier and an LLM-based scorer. That gives us a per-file, per-region view: instructors get highlighted ranges with a likelihood, so they can have a real conversation with a student instead of waving a black box. We also run plagiarism detection (Dolos-style similarity across submissions) in the same pipeline. One integrity result: plagiarism likelihood, AI likelihood, and exact line ranges for both. That’s the kind of tool we wished our own courses had.

The Bigger Problem We’re Solving

Universities still rely on outdated coding platforms that are slow, unintuitive, and frustrating. Students fight poor UIs and unclear errors; instructors fight weak grading tools and integrity that misses real misuse. Kernex is a web-based coding platform built for university CS courses: a high-performance, browser-based environment where students write, run, and debug code with immediate feedback and a clean interface.

For Students

• Browser-Based Environment: Files, terminal, and assignments live in one place.
• Real-time Collaboration: Sandboxed sessions so you can work together.
• Modern Interface: Built for 2026, not 1999.

For Instructors

• Rubric-Aware Grading: Logic that fits your course expectations.
• Academic Integrity: AI detection (KernexCheck 1 inside KGrader) plus plagiarism (Dolos), with line-level highlights and clear likelihoods.
• Scalable Infrastructure: Less time debugging student laptops; more time teaching.

We’re just getting started. Kernex aims to be the infrastructure that powers how coding is taught, assessed, and learned—with AI code detection that actually ships.