Practical coding shows up when a program runs. A paper exam asks about code, and a lab exam asks for code that works. AI tools that write programs on request have widened the gap between the two.

A coding platform for schools can close that gap with three habits:

  • On-screen exams: code that runs and is checked
  • Short spoken checks: students explain their own work
  • Small tasks: a steady rhythm through the term

Each one is explained below.

What Can a Paper Exam Not See?

Real coding has four actions: writing, running, reading error messages and fixing. A paper exam covers the first and part of the last. Nothing runs, so no error message ever appears.

  • A student who explains a loop well on paper may freeze at a blank editor
  • A student who builds well may lose marks for a misplaced bracket on a page

A coding platform for schools can test both skills in one sitting.

What Did England Learn About Protecting a Task?

A task done at home or over several weeks is hard to protect. A task done in the room, on a screen, is easier to trust. England's GCSE in computer science shows why:

  • 2017: exam boards released the programming tasks on 1 September, and solutions and advice spread online soon afterwards
  • Regulator's view: Ofqual was not confident that exam boards could detect every case
  • 2018 and 2019: the task stayed on the course but no longer counted
  • From 2022: programming skills are assessed only under exam conditions

A coding platform for schools that runs exams in the room, on a screen, fits that lesson.

What Did AI Tools Change?

Take-home programs are harder to trust now that AI tools write code on request. Two 2026 sources show why:

  • OECD: the Digital Education Outlook 2026 notes that students using general-purpose AI often produce better work. The gains tend to shrink or reverse once the AI is taken away
  • Utah State University: a Python course with 96 students allowed free AI use on weekly assignments. The share of characters entered by paste rose from 61.0% in an earlier AI-free term to 68.1%

The Utah course ran at a university, so its numbers come from university students. A school can still borrow the idea. A school policy can also label each task:

  • AI allowed: tests prompting and checking
  • Room only: tests unaided skill

Labels remove guesswork for students and parents.

Five Ways to See Real Skill

No single method sees everything. Each answers a different question:

MethodWhat It ShowsWhat It Leaves OutTeacher Time
Written paperSyntax, concepts and tracingWhether code runs, and how a student fixes itLow
On-screen lab examWorking code, checked against expected outputThe route a student tookLow, with automatic checking
Take-home projectPlanning, design and persistenceWho wrote the codeHigh
Fifteen-minute code reviewUnderstanding of submitted workThe amount of working codeHigh per student
Live coding conversationReasoning while buildingLong projectsHigh per student

Pairs of rows cover each other's gaps. A coding platform for schools can keep the lab exam and the submitted work in one place.

Can Fifteen Minutes Show Understanding?

Yes, the Utah course suggests. It let students use AI freely. It then asked each one to sit a 15-minute interview within 48 hours of every submission.

  • Weight: interviews counted for 70% of each assignment grade
  • Student view: 81% agreed that interview grades reflected their understanding. That beat quizzes and exams
  • Exam scores: proctored scores stayed in line with earlier terms, with no statistically significant difference

A class of 40 would need ten hours for one round. A draw keeps it light:

  • Five students per task sit the interview
  • The rest write two lines on screen, explaining one line of their own code
  • Over a term, every student sits at least one interview

A Live Conversation About Code

A conversational exam described in a January 2026 paper asks students to write code live while explaining their reasoning. Students may consult documentation and use supervised AI. The authors ran it with 58 students in small groups over two days. They concluded that oral exams can scale to typical class sizes.

A school can borrow the shape:

  • Three students at a time sit with the teacher for ten minutes each
  • The rest of the class works on a lab task
  • The talk opens with a request to show the program running
  • It then moves to a change: make it do this instead

Hidden Tests and Partial Credit

A student can write a program that prints the right answer for the one example in the question. It may break on any other input. Hidden test cases catch this.

  • Hidden inputs: the platform runs the code on extra cases the student never saw
  • Examples of extra cases: an empty list, a negative number or a very large value
  • Visible examples: these show what the task asks
  • Partial credit: a program that passes three of four checks earns three quarters of the marks
  • Feedback: the student sees which check failed

A copied solution may pass the visible example and still miss the hidden ones. A coding platform for schools can run those hidden checks in seconds.

What Does a Lab Question Look Like?

A good lab question is short, runnable and checkable. Take a Grade 8 example:

  • Task: write a program that finds the largest number in a list
  • Visible check: the list 3, 9, 4 returns 9
  • Hidden checks: a list with one number, a list of negatives and an empty list
  • Marks: one for each check passed, plus one for a short comment explaining the idea

Students see their score and the failed check the moment they submit. The teacher reads only the comments.

Should Students Look Things Up?

Working programmers look things up all the time, so an exam can allow it:

  • Open reference sheet: one page of commands, identical for every student and handed out with the paper. Marks then go to logic and debugging
  • Closed book: suits a check on recall of commands

The choice depends on what the exam is for.

Starting Before Text Code

Practical assessment does not need text code:

  • Scratch Jr: a young student shows a sprite that walks, turns and stops. The teacher watches for thirty seconds and notes whether it matches the task
  • Scratch: block coding carries the idea to older beginners. The stage shows at once whether a script works
  • micro:bit: a program proves itself on the board, where the lights flash or they do not

A short live check captures that moment, and a paper question cannot. These tasks give younger students a fair first taste of practical assessment. A coding platform for schools should welcome those students too.

How Should Students Prepare for a First Lab Exam?

A first lab exam feels strange, since students type under a timer while a teacher watches the room. Three habits help:

  • Practice paper: one on the same platform a week earlier. Students learn to submit work, read feedback and see a partial mark
  • Shared marking checks: the teacher hands them out in advance, so the result reflects coding and not surprise
  • Fair timing: typing speed varies widely. Size each task for the slower typists, or give every student the same extra minutes

A Term in Four Rhythms

A mix of methods keeps marking light and the picture complete:

RhythmTaskMarked ByWhat It Shows
Every weekA short coding taskThe platform, automaticallySteady practice
Mid-termOne on-screen lab examThe platform, then teacher adjustmentsWorking code under time
After each projectA five-minute talk or a 15-minute reviewThe teacher, on a drawUnderstanding
End of termA project with a spoken explanationThe teacherDesign and persistence

One possible weighting gives weekly tasks 20%, the lab exam 40%, reviews and talks 20% and the project 20%. Weekly tasks are the easiest rhythm to hand to a coding platform for schools. The checks run without a teacher. The lab exam carries the most weight because it shows working code under time.

A written paper keeps a place for ideas such as tracing a loop by hand. It simply stops carrying the whole judgment.

How Cyber Square's AI Cloud Lab Runs a Lab Exam

Cyber Square’s AI Cloud Lab serves as a coding platform for schools. Its exam features follow the points above:

  • Smart AI Teacher: the Exam Creator and Question Paper generators instantly generate practical coding questions and hidden test cases
  • Lab exams: students sit online lab exams in real time, with automated evaluation
  • Code marking: the lab marks all student code and lab assignments in one click
  • Checking: a student’s program runs during the exam, and its output is compared with the expected answer
  • Adjustments: teachers can change results before they reach the scorecards
  • Feedback: students receive instant feedback and performance reports
  • Textbooks: our computer science textbooks pair with the lab for automated homework and exam evaluation

A wrong answer turns into a lesson while the exam is still fresh in the student’s mind.

Turn a Past Question Into an Online Exam

Pick one coding question from a past paper and send it to us. We will turn it into an online exam question for your class. Then we will show how online exams work, from first submission to scorecard.