Bridgework

Inspiration

I grew up in a province in Cambodia, not in Phnom Penh, and I spent most of high school counting down to the end of it. I thought I hated studying. What I actually hated was studying with no answer to the question "what is this for", and never being given one.

My school followed the national curriculum. That means textbooks written in the early 2000s that are still in classrooms today, and a method of copy, memorise, repeat. Public schools are underfunded, private schools are expensive, and private schools with an overseas curriculum are out of reach for almost everyone. So the quality of a Cambodian child's education tracks their parents' income, and the one thing that is supposed to break that cycle ends up reinforcing it.

Where I am from, a lot of kids leave school before eighteen and go into minimum wage work. From inside that decision it is not laziness. Five more years of something that seems pointless is a bad trade against earning money now.

My siblings are in high school right now and they are in the same position. They want to build skills and there is nothing there for them. My brother could not tell you what an AI engineer does, or what someone in IT actually does beyond "writes code". Bridgework is what I wish had existed when I was fifteen.

What it does

Bridgework starts from the syllabus students are already stuck with, instead of asking any ministry to adopt a new one.

A student picks their grade, subject and topic. Grade 9 Mathematics, Trigonometry. The site shows them the real world applications of that exact topic and lets them try one as a short job simulation, with a real brief, real numbers, and three decisions a practitioner actually makes. It is marked the way a supervisor would mark it, not the way a quiz would, and finishing one unlocks the career behind it: what the day involves, how people get in, what it pays, and whether it needs a degree. Most of them do not, which is the most useful thing a fifteen year old in a province can find out.

There are simulations from Grade 7 to Grade 12. Ordering grass for a garden with a paved corner cut out of it. Working out a roof angle for solar panels. Deciding whether an AI model that is 91% accurate overall and 74% accurate for countryside students is fit to launch. The Grade 7 one has students build a working program out of coloured blocks to make a character draw a square, which is where they find out the corners are 90 degrees because four turns have to add up to 360.

The other half is that instructors publish these. University students and people doing the work write up a decision from their job, tag it to the school topic it uses, and it appears in the student catalogue immediately.

How we built it

Next.js and React, deployed on Vercel, with no backend. Content is stored in the code, and accounts, progress and instructor submissions live in the browser. Marking runs on the device with no model call, so a student on a weak connection gets the same feedback instantly.

I am a one person team, and I built this with AI assistance (Claude) throughout, inside the hackathon window. I want to be direct about that. I could scope the product, map it to the curriculum, write and judge the content, and tell when something was wrong. I could not build a backend on my own. What is submitted is an AI assisted prototype of a working idea, not a production system, and I would rather say that than imply otherwise.

Challenges we ran into

Most of the difficulty was in thinking, not typing.

For this to reach the students it is for, it has to be in Khmer, and it has to work offline. Neither is true yet. The marking already runs on the device, so offline is reachable, but I did not get there tonight. In the meantime the more realistic route may be teachers rather than students. A teacher can take these and make their own lessons less abstract, without every child needing a phone.

The rest is about getting it into schools at all. What a partnership with a school actually looks like, who writes the simulations at scale, and whether this works outside Cambodia. I think it does. Students in other developing countries face the same curriculum problem, and even in Australia there are plenty of students sitting in class wondering why any of it matters.

Accomplishments that we're proud of

The whole path works end to end, solo, in one evening. A student signs in, picks a topic from their own syllabus, does the work, gets marked, and unlocks a career. An instructor publishes a simulation and it appears for students straight away.

I am most pleased with the marking. I assumed I would need a language model to grade written answers. Writing the rubric by hand turned out to be faster, works with no connection, cannot fail during a demo, and cannot invent a mark. Being out by a factor of ten is reported as a unit conversion problem rather than just "wrong", and a block program tells you how many steps ran correctly before the first mistake.

I am also glad I caught two things late rather than never: a simulation written in language nobody could follow, and a hole that let people answer without an account.

What we learned

SDG 4 is quality education, and I used to read that as getting children into classrooms. Building this changed what the word quality means to me. These students are already in classrooms. What they are not given is any reason to stay.

Accessibility turned out to be broader than I expected. For the students I am building for, the barriers are where they live, what their family earns, and what language the material is in, long before anything else.

The clearest lesson came from a simulation I wrote and then threw away. My first Grade 9 Pythagoras task was about cutting roof rafters, and it used words like span, ridge and wall plate. It was accurate and completely impenetrable, and I am at university. I had rebuilt the exact thing I was angry at: content that is hard because of its vocabulary instead of its ideas.

So I rewrote everything to one rule. If a word would send a student to look it up, it does not go in. The replacement is a person at a window six metres up, and a ladder whose base has to sit 1.5 m out from the wall to be safe:

$$L = \sqrt{6^2 + 1.5^2} \approx 6.18 \text{ m}$$

Same maths. The 6 m ladder does not reach. Nobody needs a glossary for it.

What's next for Bridgework

The content is the product, so the next step is people. I want to build a network of contributors: working professionals, and final year university students who have just been through the pathway and can still remember what they did not know. More voices means more subjects, more careers, and simulations that are more fun than anything I can write on my own.

After that, make it richer to look at. Images and short video walkthroughs of what a job actually looks like, since a photograph of a surveyor on a road does more than a paragraph about one. I also want to link out to free places students can keep going, starting with Scratch for anyone who liked the coding task, so a simulation becomes the beginning of something rather than ten minutes.

Then the two things I already know are missing: Khmer throughout, and full offline support. And after that, schools. Into Science already sends university students into secondary schools here, and that is the model I want to work with rather than around.

Built With

• and-accounts
• and-no-external-apis.-content-is-stored-in-the-code
• claude
• css
• deployed-on-vercel-from-github.-no-database
• git
• github
• google-fonts
• html
• javascript
• next.js
• no-backend
• node.js
• progress-and-instructor-submissions-use-the-browser's-web-storage-api.-marking-logic-is-plain-javascript-running-on-the-device.-fonts-are-fraunces
• react
• styled-with-hand-written-css-using-custom-properties-for-theming
• svg
• vercel
• web-storage-api