Learn to code on any device. No computer, no set-up.

Rich or poor, most students have nowhere to practise code

1. No computer, or none they can use. Many families have no computer at home. Some students only have an iPad, often one subsidised through school. Other parents choose not to buy a computer at all, because without the know-how to make it a learning tool it simply becomes a games machine.
2. Set-up is a wall. Installing and configuring a programming environment is hard even for families who can afford a computer, and many give up before the first program runs. The step from wanting to try to seeing code run is far too steep.
3. Exam questions are mostly blank. HKDSE programming questions do not ask students to fill in one word. They leave a large part of an algorithm empty. Facing a nearly blank block, many students freeze, and a marked paper never shows where their thinking stopped.
4. Drilling is not learning. Many learning apps are endless multiple choice. Students never get to change code, run it and see what happens, which is how programming is actually learned.
5. Robotics needs hardware. Boards, sensors and tools cost money and set-up time, before a student even knows whether they enjoy it.

Prototype · used with real students

A coding page that follows how students think

Students open an HKDSE-style question in the browser, on an iPad or any device, and write the missing part of the program right there. As they type, the AI reads what they are writing, works out why they are thinking that way, and gives a hint that fits: where to begin, what to try next, or what a change will do. They can edit and run the code freely, experiment, and see the result for themselves.

def count_above_average(scores):
    # ---- complete this part ----
    total = 0
    for s in scores:
        total += s
    avg = total / len(scores)
    ▌
    # -----------------------------
    return count

In the prototype today

Zero set-up

Python runs inside the browser on any device, including an iPad. Nothing to install and nothing to configure, so it is just as useful in a home that does have a computer.

Real exam shape

Questions follow the HKDSE format, where a large part of the program is left blank, not one-word fill-ins.

Hints that follow their thinking

As the student writes, the AI infers the reasoning behind their code and responds to it: where to start, what is missing, or why a line behaves as it does. Never the full answer.

Experiment freely

Change any line, run it, and step through to watch the variables. Prompts like what happens if you change this? make cause and effect visible.

Graded difficulty

Questions at several levels, so every student starts where they can succeed and climbs from there.

Tutor report

Shows whether a student had no idea where to start, stalled at the first step, or hit one specific bug, so the tutor knows exactly what to explain next.

Next

Practice from homework

After a student finishes an assignment, such as an HKDSE ICT paper, the engine finds the concepts behind each mistake and writes similar questions that target them, reviewed by the tutor before they are assigned.

Physics

The same diagnose-and-practise loop for HKDSE Physics.

Why it matters: this is not another multiple-choice drill. Students learn programming the way it is really learned, by writing, running, breaking and fixing code, with help at the exact moment they are stuck, on whatever device they have.

Next

See what the hardware is doing, before you buy any

Existing simulators can blink an LED on a virtual board, but they rarely explain what is happening. Ours shows it: wire up a board, a sensor and a motor in the browser, run your code, and watch every signal travel between them, with a plain-language explanation of each step.

What each part does

The board

The brain. It runs your code and decides what happens next.

Sensors

The senses. They measure the world, such as temperature, distance or light, and report numbers back to the board.

Motors and lights

The hands. They turn your code's decisions into movement and signals you can see.

The wires between

The conversation. Parts talk using agreed rules, like a shared language. The simulator shows each message as it is sent and explains what it means.

Try it first, understand every part, then take the same code to a real board.

Live

Taught by someone who has been through it, and done well

• 4 × 5**2023 HKDSE: Core Maths, M2, Physics, ICT
• 4.05 / 4.3Cumulative GPA at HKUST, Dean's List every semester
• 3Years in a row invited back to teach a Band 1A school's HKDSE ICT elite class
• 10+Private students tutored, from secondary school to university

HKDSE

Core Maths · M2 · Physics · ICT

University level

Maths, physics and programming courses

Beyond the syllabus

Programming, competitive programming, embedded systems and robotics

Teaching

• Invited back three years in a row to teach the HKDSE ICT elite enrichment class at a Band 1A secondary school, preparing its top students for the exam. Every student's results improved, and the class was very well received.
• In charge of a 100-person embedded programming tutorial at HKUST, and co-led a training programme for about 250 participants.
• Three years of private tutoring, one-to-one and in small groups, from HKDSE to university courses.

Secondary school

• 5** in all four of Core Maths, M2, Physics and ICT in the 2023 HKDSE.
• Competed in the Hong Kong Physics Olympiad, the IMO preliminary selection and the Hong Kong Olympiad in Informatics.

University

• HKUST Technology and Management dual degree: BEng Computer Science (Extended Major in AI) and BBA. Computer Science GPA 4.27 / 4.3.
• Gold Medal and 11th place at the 2024 ICPC Asia Hong Kong Regional Contest.

AI handles the practice. Tutors do the teaching.

There is no point spending a lesson watching a student type out one programming question, and most of that typing teaches little on its own. With ZeroGap, the practising, the trial and error and the first round of hints happen on the platform. Lesson time goes to what only a tutor can do: explaining the ideas a student still does not understand.

Built from real teaching

Now

Prototype with zero-set-up coding, step-by-step execution, in-context help and tutor reports, in private use with tutoring students, with positive feedback.

Next

Question generation from finished assignments, and a pilot with partner tutors and schools, starting with HKDSE ICT.

Later

HKDSE Physics and the Robotics Sim Lab.

Ivan Lok

I have taught students from all kinds of families, rich and poor, and many of them had the same problem: no computer they could actually code on. Some had only an iPad; others had a computer that was never set up for learning. All I could give them was an online compiler and a lot of copy-and-paste, and I watched capable students give up on the tool, not the subject. ZeroGap started as the tool I wished I could hand them.

• Builds the product himself: research experience in LLM-guided code generation, and full-stack and embedded engineering.
• Years of hands-on embedded and robotics engineering, leading software teams and writing the hardware libraries the Sim Lab will simulate.

Full profile and CV →