Loading…
Launch notice. CodeTeach.ai is in its initial production release. Please report any issues to admin@codeteach.ai and generated assignments should be closely reviewed for quality and accuracy.
Loading…
Quick answers below, or browse the full Help Center for in-depth articles.
Step-by-step guides for every part of the workflow — picking a generation mode, choosing a difficulty level, the autograder check, deploying to GitHub, and troubleshooting.
Search and browse 20+ articles organized by collection.
AI creates candidate artifacts. Trust comes from the checks that follow: observed program behavior, isolated test runs, complete-package review, and an autograder proven in GitHub.
CodeTeach uses the generation mode to decide what the instructor owns and what AI may create.
The app does not trust an AI model to guess the exact output a program should produce.
Before Review, deterministic checks confirm that the artifact set is complete enough to grade.
After the instructor reviews and approves Instructions.md, CodeTeach validates the current package as one unit.
Deployment has a separate technical gate because a sound assignment can still be packaged into a broken workflow.
CodeTeach does not replace instructor judgment. It automates the repetitive, cross-file, and environment-specific verification that is difficult to perform consistently by hand.
A fluent AI response is not treated as proof. Code, tests, and the autograder must run.
The same isolated validation can be rerun after edits, avoiding the easy-to-miss differences of an informal manual check.
CodeTeach compares artifacts that are often reviewed separately: what students are told, what the solution does, what tests enforce, and what the rubric scores.
Instructor-provided work is not silently changed to make a check pass. When intent is unclear, the issue returns to Review.
The autograder is checked internally before publication and again in the real GitHub Actions environment afterward.
Skipped AI reviews, warnings, failed tests, and post-push failures remain visible. CodeTeach does not turn missing evidence into a clean result.
You remain the authority on learning objectives, difficulty, accessibility, institutional policy, and whether the assignment is appropriate for your students. Read the instructions and inspect the artifacts before approval; CodeTeach supplies evidence and catches inconsistencies, but it should not make the pedagogical decision for you.
Have feedback or found a bug? We would love to hear from you.
Or email admin@codeteach.ai.