Visual Quiz: keep the answer
in sign.

A multiple-choice question, four labeled choices, and one change to the oldest test format on earth: you answer by signing A, B, C, or D to the camera — no clicked button, no thumb on a neighbor's screen, the response kept in the same language the lesson was taught in.

What Visual Quiz is

Visual Quiz is the most common assessment format on earth with one channel swapped. The question is shown on the screen. The four choices are labeled A, B, C, D. The student answers by fingerspelling the letter to the camera instead of clicking it. A short hold-to-confirm window keeps a half-formed handshape from counting. At the end, there's a score, and an embedded quiz can pass that score back to the host page or a gradebook.

The reason it exists is a small gap that repeats. A Deaf student takes the lesson in sign. The videos are captioned. The instructor signs. Then the test hands them a mouse. None of that is hostile — it's just an afterthought, the response channel treated as the one part nobody designed. Visual Quiz closes the gap by keeping the answer in the language the learning happened in. The question is visual; the answer is signed; the whole loop stays put.

It comes through two doors, because two kinds of people ask for it. A teacher or a course author wants to build a short quiz and drop it onto a page they already run. A museum or a district wants the whole thing built around their content and wired into their floor. Same engine underneath, two ways in — and the access floor is the same on both: a working keyboard answers A through D, captions are on by default for anything spoken, and the student's video never leaves the browser.

What we built

The hard part of a sign-answered quiz isn't the question. It's making the answer trustworthy — a letter the camera read on purpose, not a flicker it guessed at — and making the whole thing degrade to something a kid without a camera can still take. Both of those are in the engine, not bolted on per quiz.

• Sign-to-answer The student fingerspells A, B, C, or D to the camera. A hold-to-confirm window means a half-formed handshape doesn't count — the letter has to hold steady before it registers, so a nervous hand or a passing gesture can't answer for you. The recognizer is trained on Deaf-signed handshapes, not a hearing approximation of them.
• Keyboard fallback, by design No working camera, or a student who'd rather not be on one, presses A through D and takes the exact same quiz. The fallback isn't a downgrade path you hit when something breaks — it's a first-class way in. Same questions, same scoring, same record.
• Runs in the browser The quiz runs in the page; there's no server-side processing of the student's video. The camera feed stays on the laptop. What leaves the browser is a score, and only if the host page asked for one — the picture of a child's face never does.
• Score passback At the end the student sees a score. An embedded quiz can post that score back to the host page or a gradebook, so a teacher gets the number where they already keep their numbers — no second system to check, no roster to reconcile by hand.
• The self-serve embed One frame, camera permission, a quiz ID. It drops onto any page that allows a camera iframe and it answers the same way everywhere — a class site, a course module, an exhibit microsite. No studio build required to get a working quiz on a page.
• Quiz Lab commissions When a quiz needs to be part of an exhibit or a broadcast, the studio builds it: the content scoped to your curriculum, your visual treatment, the integration into your kiosk or gradebook, and the access floor — keyboard fallback, captions on — written into the build rather than hoped for.

* Projected. One classroom pilot is the near-term target for the first term — the self-serve embed and the demo are live today; the gradebook-passback and white-label tiers roll out as schools come on.

The same classroom, the same way

Here is the picture the whole product is built around. A classroom with a Deaf kid and a hearing kid in it, both taking the same quiz, the same period, off the same engine. The Deaf kid signs the answer. The hearing kid presses a key. Neither one is on a special track, and neither one is waiting on the other — nobody got handed an accommodation form, nobody got pulled into a separate room with a separate test, and nobody bolted accessibility on at the end because the test was built that way from the first question. That's the whole idea, and it's the part that's easy to say and hard to actually ship.

Visual Quiz answers a question that's older than any of the technology: how do you run one assessment that a signing student and a speaking student both take on equal footing, in the same room, at the same time? The answer the product lands on is to design the response channel first instead of last — to treat signing your answer as a real way to answer a test, not a courtesy. A multiple-choice question doesn't care how the letter arrives. We made the letter arrive in sign.

If the lesson was in ASL, the test should be too. The day a signed answer counts for less than a clicked one, we built it wrong. — a working principle, written into the engine on day one

Answer one.

Below is a small, working slice of the quiz: one question, four choices. Sign your answer in the demo above with a camera, or right here, pick a letter the way the keyboard fallback does it — tap a choice, or press A, B, C, or D on your keyboard. The hold-to-confirm bar fills, then the answer counts and scores, exactly the way it does when the letter comes in from the camera. Nothing here phones home; it's the answer-and-score loop, running in your browser.

Why this fits 1891

The pitch on madeby1891.com is for systems designed so the access is the floor, not a feature — built by Deaf people, built for everyone in the room. Visual Quiz is that pitch applied to a test. The answer is given in sign because sign is a real way to answer. The keyboard path is first-class, not a fallback you're sorry to offer. The video stays on the laptop because a child's face is not data anyone needs to keep. None of that is an accessibility skin painted on at the end — it's the order we built the thing in.

There's a longer line behind it, too, and it's quiet on purpose. The classroom where a Deaf kid and a hearing kid learn side by side is not a new idea here. The Deaf school, the signed lesson, the test that finally fit — that's the lineage Visual Quiz belongs to. A signed answer counting for full credit is a small thing and an old want. We just built the part where the test agrees.

What's still unfair about this

The promise we made the recognizer is that a signed answer is exactly as trustworthy as a clicked one — no more wrong reads, no penalty for answering with your hands. That bar is the product. The day we'd take a faster recognizer that's a little less sure, we'd be quietly telling signing students their answers count for less, and that's the one trade we don't make. We tune the camera until the hand is as reliable as the mouse.

The thing we did not do is treat the camera as the only door. A student who can't or won't be on one presses A through D and gets the same quiz, the same score, the same record — not a lesser version flagged as an exception. And a teacher who outgrows the free tier can export their scores and walk; there's no retention trick holding a gradebook hostage. We measure Visual Quiz by whether a signed answer and a typed answer end up indistinguishable in the grade, not by how clever the camera is.

Nobody taking the quiz cares about the recognizer. They care that the answer they gave is the answer that got counted. — Anthony Mowl, framing the build for Visual Quiz

That sentence is the spec. The recognizer is good because the student never has to think about it.