Branch · 01
Engineering branch
Low-level hardware-software architecture
Products
AI4SGiftedInnovatorProgram
With AI4S as the knowledge direction and the Human–AI Double Helix as the creative method, spanning courses and workshops,linking camps, competition incubation and a portfolio archive — from “using AI” to “creating with AI.”
Human questionAI generation
S-P-M
Growth ladder
5 STEP
Inquiry loop
4 ASSETS
Sci-tech assets
Why AI4S
Large models are rewriting the logic of education.
When AI can generate quickly, simulate and assist execution, the truly scarce ability is no longer repeating standard answers — it's asking questions, judging value, orchestrating tools and creating works.
Since 2023, UNESCO has continued to examine the impact of generative AI on education, releasing a series of studies and guidance documents that urge educators to rethink curricula, ways of learning, assessment systems and people's core competencies.
July 2023 — UNESCO published “Generative AI and the future of education.”
September 2025 — it released “AI and the future of education: disruption, dilemmas and directions.”
UNESCO Questions
01
As AI keeps surpassing humans in more fields, what should future education look like?
02
Which skills, mindsets and abilities should education cultivate?
03
How should curricula and teaching methods adapt to the times?
04
How should the teacher's role, students' ways of learning and assessment systems change?
05
When AI can answer most questions, what abilities do children still need?
Core Method
Human questioningAI's generative power.
We don't train skills that “get replaced by AI” — we teach children to command AI: ask questions, define boundaries, organize knowledge, orchestrate tools and turn ideas into real works.
Deliverables
Visible personal sci-tech assets.
A leap from hobby classes to a hardcore sci-tech portfolio. Every child who leaves MicroRealm takes away not a certificate, but a continuously growing set of digital hard currency.
Asset 01
Full-stack self-built apps
Moving from rote syntax to understanding the “whole architecture,” using Vibe Coding to quickly build software apps that solve real problems.
Example
e.g. a campus plant-ID web page, an AI mistake-book assistant, a science-experiment logging system.
Growth archive materials
App prototypePage structureFeature notesIteration log
S-P-M Architecture
The S-P-M architecture: a growth ladder toward future top-tier innovators
From perception, expression and observation in the early grades to engineering practice, AI collaboration and project creation in the later grades — forming a continuous growth path.
PHASE 01
S Phase(Grades 1–4)
S1
Begin the sensory journey：Early visual/auditory recognition interaction to build perceptual understanding.
S2
Explore intelligent logic：Learn basic conditional logic and experience the concrete beauty of rule-making.
S3
Model-training first steps：Graphically experience how AI captures features, sensing the leap of defining aesthetic rules.
S4
IoT sensing & feedback：Build sensing art installations and grasp core concepts of data collection and hardware feedback.
PHASE 02
P Phase(Grades 5–7)
P1
Edge engineering practice：AI vision cameras and edge computing — establishing the engineering ability to map compute onto the physical world.
P2
Full-stack development：Use Vibe Coding for full-stack web app development, master code debugging, and gain the ability to define the boundaries of complex systems.
P3
Mastering AI applications：Deploy large models and build personalized agent projects. The goal here is not “hand-writing lots of code,” but “understanding system structure, reading key logic, and using AI to complete runnable prototypes.”
PHASE 03
M Phase(Grade 8+)
Multi-agent collaboration & π-shaped strengths：Core role: as a full-stack agent coordinator (Harness Engineer), students orchestrate clusters of agents to tackle real industry pain points, benchmarked against Silicon Valley's frontier agentic workflows. Acting as the “chief engineer (Harness Engineer),” they schedule multiple AI experts inside a highly encapsulated research sandbox to solve complex problems.
Teaching axis：Multi-agent orchestration, RAG knowledge-capsule injection, and cross-disciplinary game-theoretic reasoning. Benchmarked against Silicon Valley R&D standards, anchored on RAG truth points, with mandatory game logs.
Core value：Focusing on real pain points across new agriculture, engineering, medicine and liberal arts to forge π-shaped talents with industry insight and compute-orchestration ability.
Capability Matrix
Full-stack capability framework matrix
Dimension
AI tech literacy (Tech)
General / disciplinary / arts enrichment (Enrichment)
01Logical thinking
AI tech literacy (Tech)Algorithm design, nested conditionals, multi-agent workflows
General / disciplinary / arts enrichment (Enrichment)Introduces formal logic and mathematical modeling.
02Engineering ability
AI tech literacy (Tech)Full-stack development, visual architecture, agile testing
General / disciplinary / arts enrichment (Enrichment)Introduces engineering management, industrial standards and fault-tolerance design.
03Orchestration engineering
AI tech literacy (Tech)Prompt design, context management, and building AI multi-agent engineering environments
General / disciplinary / arts enrichment (Enrichment)Introduces rhetoric, logical semantics and domain-literature reading.
04Aesthetic literacy
AI tech literacy (Tech)UI/UX design, interaction aesthetics
General / disciplinary / arts enrichment (Enrichment)Introduces algorithmic aesthetics, visual narrative and structural-mechanics beauty.
05Ethics & doing good
AI tech literacy (Tech)Data privacy, on-device models, anti-hallucination mechanisms
General / disciplinary / arts enrichment (Enrichment)Introduces tech ethics, copyright awareness and digital sovereignty.
Learning Loop
The five-step AI self-inquiry method.
Courses aren't about “finishing the syllabus” — children complete works through a loop of prompt, self-study, discussion, practice and reflection.
STEP 01 · 命题
Prompt
The teacher sets a challenge — a theme or a real problem to solve — with an AI agent assisting task comprehension.
Competition Pathway
Annual sci-tech
competition navigator.
Competitions aren't the goal — they're a milestone check within the scientific research process. We care more about topic fit, ladder-like growth and differentiated competition.
- Daily incubation
Original topic incubation & development
Around “one student, one topic,” complete initial modeling, data collection and core validation — building ability first, not competing for the sake of competing.
- Milestone validation
Milestone validation contests
Use whitelisted specialized contests to test the depth of semester topics, build children's confidence and accumulate material for their résumés.
- Project incubation
Sci-tech talent contests · incubation phase
Complete formal applications, polish works and practice defenses — turning the learning process into verifiable project outcomes.
- Showcase
Sci-tech talent contests · showcase phase
Showcase a year of growth in on-site challenges and pitch defenses, accumulating certificates, experience and long-term sci-tech background.
Q&A
Q&A
Q&A07 Questions
A1
No. At a young age it's not about complex low-level code and architecture, but about building basic understanding of AI through observing life, expressing ideas, experiencing interaction and hands-on activities. The form of works also progresses with age: younger grades start with observation, expression, simple interaction and creative works; middle grades gradually complete software-hardware and AI application projects; advanced levels form research directions and portfolios.