A 92-second video from Shanghai's WAIC 2026 shows a presenter writing y = x² - 3, opening the corresponding parabola, sketching a rough shape, and turning a flat drawing into a labeled, rotatable 3D geometry model. The clip crossed 124,000 views on X by the time explainx.ai verified its metadata on July 23.
The source post by journalist Jason Smith calls these "the next generation of school blackboards in China." That is directionally fair. But replies quickly inflated a conference booth demonstration into a broader claim that classrooms across China already use this exact experience by default.
Here is the evidence-backed version: the product matches iFlytek's Tongchuang AI Blackboard (讯飞同窗AI黑板), demonstrated at the 2026 World Artificial Intelligence Conference. iFlytek reports meaningful deployment across China, but the viral video alone proves only that the system worked during this public demo. We watched the full clip, checked the official product description, and separated interface capability from education outcomes.
TL;DR: What is real in the smart-blackboard video?
| Question | Evidence-backed answer |
|---|---|
| Where was it filmed? | A live booth demonstration at WAIC 2026, Shanghai |
| Which product? | Functions and interface match iFlytek Tongchuang AI Blackboard |
| What does it do? | Handwriting capture, equation graphing, sketch recognition, 2D-to-3D geometry, resource retrieval |
| Is the clip edited CGI? | It appears to show a real interactive display responding live; no independent lab test was performed |
| Is it in every Chinese classroom? | No evidence for that claim |
| Is it deployed at scale? | iFlytek reports 33 provincial-level regions and 1,400+ counties/districts |
| Is every feature generative AI? | No; AI recognition can hand off to deterministic graphing and 3D tools |
| Does it improve learning? | Plausible for spatial concepts, but the video is not an outcomes study |
| Can schools buy it globally? | Availability, pricing, curriculum mapping, and support vary; no global retail price is established by the clip |
Watch the full WAIC 2026 AI-blackboard demo
Full 92-second clip from Jason Smith's July 23 X post. Locally embedded as WebM for reliable playback; original source and attribution preserved.
The sequence matters because several retellings compress different functions into "AI creates 3D worlds from chalk." What the presenter actually demonstrates is more grounded:
- She writes
y = x² - 3on the digital surface. - The system opens a graphing tool and plots the expected upward parabola.
- She draws a rough planar shape using the stylus.
- The software recognizes or standardizes the geometry.
- A labeled 3D object appears and can be rotated with touch.
- Faces, edges, or regions can be selected and recolored.
- A related content card opens beside the object.
The last seconds briefly show another immersive visual before the presenter returns to the board. The clip does not show a full lesson, a student assessment, or long-term classroom use.
How we identified the product
The strongest match is iFlytek's official July 17 page, "iFlytek at WAIC: using AI to unlock educational productivity". It names the Tongchuang AI Blackboard and describes a nearly identical floor demo: a visitor draws a triangle, the system completes it as a standard shape, recognizes the knowledge point, surfaces related material, and converts a planar figure into a 3D model that can be rotated to inspect points, lines, and faces.
That is more specific than visual logo guessing. It ties the observed actions to a named exhibitor, product, date, and venue. The official WAIC site confirms the conference ran July 17–20 in Shanghai, while a Shanghai government event summary says the exhibition covered more than 100,000 square meters, involved 1,100+ companies, and included more than 300 global product debuts.
The white-shirt presenter also appears at a purpose-built classroom display with multiple writing panels, an integrated camera, multi-touch modes, and iFlytek-branded material. Still, the article's attribution rests on feature and event-source correspondence—not on a blurry shirt logo.
What part is actually AI?
Calling the whole board "AI" is convenient marketing, but the system is better understood as a pipeline.
stylus or touch input
↓
stroke capture + handwriting / shape recognition
↓
subject and intent classification
↓
math engine, geometry tool, or curriculum retrieval
↓
interactive graph / 3D object / teaching resource
↓
lesson capture, summary, and follow-up materials
The recognition layer is where computer vision, OCR, multimodal models, and iFlytek's education-specific AI can help infer what a teacher intended. If a rough triangle is recognized correctly, the rest can be precise conventional software: computational geometry draws clean edges; a math engine plots the equation; a renderer rotates the object.
That division is a strength. A classroom does not need an LLM to hallucinate a parabola. It needs AI to remove input friction, then deterministic software to preserve mathematical correctness. The same lesson applies to AI-native calculus courseware: the model should interpret the learner's question, while reviewed content and reliable renderers anchor the answer.
Is this really deployed across China?
iFlytek's official WAIC post says the AI Blackboard is in regular use across 33 provincial-level administrative regions and more than 1,400 counties or districts. It also says the connected Spark Teacher Assistant covers 1 million teachers, while its classroom-recording analysis system has landed in more than 700 counties or districts.
Those are large figures, but readers should preserve three qualifiers:
- They are company-reported, not an independently audited census.
- "Deployed in a region" does not mean every school or classroom in that region uses the hardware.
- The footprint can include different iFlytek board generations, software configurations, pilots, and procurement programs—not necessarily the exact WAIC demo setup.
There is evidence beyond the booth that this is not vaporware. iFlytek previously documented schools using its Spark smart blackboard for intelligent handwriting, 3D geometry, resource recommendations, and lesson recording. A 2024 case page about the High School Affiliated to Renmin University reported teacher use of subject tools, with solid geometry the largest category. Those are implementation examples, not randomized proof of educational impact.
The honest sentence is therefore: iFlytek has a scaled Chinese education deployment, and WAIC showed its latest integrated workflow. "All Chinese classrooms have this" is unsupported.
How is it different from a SMART Board or Promethean display?
Interactive whiteboards are not new. Teachers have used digital ink, object manipulation, graphing widgets, screen sharing, and lesson libraries for decades. SMART's current education displays already support touch interaction, whiteboarding, and classroom apps. Promethean follows the same broad category.
iFlytek's 2026 pitch is integration depth:
| Layer | Traditional interactive board | iFlytek AI Blackboard pitch |
|---|---|---|
| Writing | Digital ink and erase | Natural handwriting plus live digitization |
| Recognition | Basic shapes, text, equations depending on software | Subject-aware recognition and knowledge-point matching |
| Visualization | Teacher opens a prepared graph or model | Board jumps from rough input to graph or manipulable 3D object |
| Resources | Search library manually | Contextual recommendations from recognized content |
| Interaction | Touch, stylus, casting | Touch, stylus, voice, virtual figures, multi-panel workflows |
| After class | Save or export a board | Structured lesson capture, micro-lesson clips, summaries, analysis |
| AI stack | Vendor-specific assistants | iFlytek Spark plus education-specific tools and data workflows |
That does not automatically make it better for every classroom. A cheaper tablet, projector, and GeoGebra can teach the same parabola. The value depends on how much setup time the integrated workflow saves and whether teachers can use it without the technology taking over the lesson.
Why the demo is educationally compelling
The strongest moment is not the AI label. It is the reduction in representation switching cost.
A teacher can move from handwritten equation to graph without walking to a computer or preparing a slide. A rough spatial sketch becomes an object students can rotate. A face can be highlighted while the teacher explains cross-sections. This makes abstract relationships visible at the moment a question appears.
That matters especially for:
- functions, where parameter changes should visibly move a graph;
- solid geometry, where students struggle to infer 3D structure from 2D lines;
- chemistry, where formulas, molecules, and reaction steps connect;
- physics, where vectors, trajectories, and forces benefit from animation;
- remote classes, where synchronized digital writing is easier to transmit;
- lesson review, where searchable board state can anchor a recap.
The viral response resembles the excitement around an AI-built 3D cell explorer. Both make invisible or spatial concepts manipulable. But a vivid interface is a teaching aid, not a pedagogy by itself. The Dartmouth AI tutor research is a useful counterweight: learning gains depend on guided practice and instructional design, not merely adding chat or animation.
What the viral video does not prove
Recognition accuracy under classroom conditions
The presenter knows the workflow and draws expected shapes on a clean surface. A real classroom brings rushed handwriting, occlusion, multiple students writing, ambient noise, mixed languages, accidental touches, and network interruptions. Schools need error rates and recovery behavior, not only a successful stage path.
Better learning outcomes
The clip contains no pre-test, control group, retention measure, or accessibility evaluation. Rotatable geometry may improve spatial intuition, but it could also become a spectacle if students do not predict, manipulate, explain, and practice.
Teacher workload savings
Automatic lesson capture and summaries could save time. They can also create review work if transcripts and knowledge tags are wrong. The meaningful metric is minutes saved per week after training and correction—not how fast a single triangle becomes a model.
Privacy and governance
iFlytek describes a 4+1 camera setup, AI audio processing, classroom recording, summaries, and behavior or interaction analysis. That makes data governance central. Schools should ask where video, voice, handwriting, student work, and inferred learning data are stored; who can access them; how long they are retained; and whether families can opt out.
Price and lifecycle cost
The viral post has no procurement price. Hardware cost is only the start: installation, network, training, content licenses, repairs, cameras, security updates, and replacement cycles determine the real total.
A practical evaluation checklist for schools
Before a district compares countries or buys a fleet, run a small evidence-based pilot:
- Give ordinary teachers—not vendor presenters—the same five tasks.
- Test messy handwriting, left-handed use, multi-touch, and several grade levels.
- Measure time from intent to usable visualization against existing tools.
- Review mathematical and curriculum accuracy with subject leads.
- Test offline behavior and recovery after network or power failure.
- Include students with visual, motor, hearing, and cognitive accessibility needs.
- Audit camera, microphone, retention, administrator, and export controls.
- Measure student explanation and delayed retention, not only engagement.
- Calculate five-year total cost per classroom.
- Require exportable lesson assets so content is not trapped in one vendor.
For teachers adopting AI more broadly, the same governance principle appears in Claude for Teachers: protect student data, ground the model in reviewed curriculum, and keep the educator responsible for instructional judgment.
explainx.ai's read
The WAIC clip is genuinely impressive because it hides tool switching. The presenter writes as if using a board; the system recognizes intent and opens the right mathematical representation. That is a credible improvement over navigating nested classroom software menus.
The most interesting product design is also the least cinematic: AI handles ambiguity, deterministic tools handle correctness, and the teacher stays in control of the explanation. If the board can do that reliably in noisy classrooms, it is more useful than a generic chatbot bolted onto a display.
The geopolitical captions add heat but little evaluation value. China's large education market and iFlytek's deployment footprint make it possible to iterate classroom AI at scale. Western vendors have decades of interactive-board experience. The serious comparison is not "China has the future, the West has chalk." It is recognition accuracy, curriculum quality, teacher adoption, privacy, accessibility, learning outcomes, and total cost.
Related on explainx.ai
- Academa.ai: LLM-integrated multivariable calculus
- Dartmouth Phosphor AI tutor study explained
- Claude for Teachers: free US K-12 access and privacy
- AI-built interactive 3D cell explorer
- AI curriculum for kids K–5
- How to teach kids to spot AI fakes
- China AI playbook: open models and cheap compute
- US vs Chinese AI startups
Primary and contextual sources: Jason Smith's X video · iFlytek's official WAIC education post · WAIC official site · Shanghai government WAIC summary · SMART education displays
Video behavior, X engagement, event context, and iFlytek product claims are accurate as of July 23, 2026. Deployment figures are company-reported; explainx.ai did not independently test the hardware or verify classroom learning outcomes.